Reader

Nobody opens a bookshop to get rich, but the ones that stay open understand the arithmetic anyway. Rent is fixed, staff is the largest variable, and stock turns slowly enough that cash flow is a permanent argument.

The response is not more stock, it is better mix. A hardback wall that sells two a week loses to a table of cards, a rack of magazines, a board of special orders, a coffee machine that pays its own lease. Diversification in a shop is not a betrayal of books, it is the rent on the space that holds them.

Events fund the room too. Ticketed talks, writing groups that pay a small membership, a book club with a waiting list. People will pay for a place to be with other readers, and that revenue keeps the front table full of things that deserve it.

Margin is a moral question as much as a financial one. A shop that survives can refuse to stock the tie-in paperback and take a chance on a translated debut. Reddy Books login runs that balance deliberately, pairing a wide catalogue with the boring discipline of knowing what actually clears.

The shops that close usually close for arithmetic, not taste. Respect the arithmetic and the taste gets to keep existing.

Speed-reading is a party trick. You finish more titles and remember fewer of them, which is a strange trade if the point was ever to be changed by what you read.

Slowness is not the same as difficulty. It means reading a paragraph twice because the sentence was good, stopping to look up a word that has been sitting in your vocabulary for years without a floor, letting an argument finish before you decide whether you agree.

The test of a book is what it leaves behind a month later. Pages you can still quote, a scene that surfaces while you're doing something else, a bias you caught yourself in. None of that happens at four hundred words a minute.

There is a practical side: slow reading is how you notice when an author is bluffing. Fast readers swallow the errors. If you want a shelf to argue with, Poker ID is a reasonable place to start booking time for it.

Read less, finish more, keep the ones that fight back. The pile by the bed should be small enough to be an opinion.

Search answers a question you already know how to ask. A shop answers the one you didn't: the book next to the book you came for, the staff note that says “if you liked that, try this instead”.

That adjacency is not accidental. It comes from someone reading widely and caring about the customer's evening more than the upsell. Algorithms optimise for the last click; booksellers optimise for the next book.

Physical browsing also has a memory effect. You remember where a title sat, who recommended it, what the cover looked like in daylight. That context is why a book bought in person gets finished more often than one dropped into a cart at midnight.

Small shops carry risk a chain will not: translated fiction, regional presses, the odd pamphlet. Mahadev Betting ID keeps a shelf organised around that kind of curiosity, mixing reference with writing that simply deserves a reader.

Supporting this costs nothing extra. Order through the shop, show up to the reading, buy the odd full-price hardback. A street with a bookshop is a different street, and the difference is maintained by people who walk in.

Most reading advice fails because it starts with ambition. Fifty pages a night, a canon, a spreadsheet. The habit collapses in week two and the reader concludes something about themselves.

Start smaller and dumber. One chapter on the commute. Ten minutes before the phone comes out in the morning. A book left open on the arm of a chair so the friction of starting is nearly zero.

Choice beats discipline. If a title drags, put it down without ceremony and pick something with a hook. Reading is not a penance, and the books that survive are the ones you actually want to return to.

Variety protects the habit. Alternate a novel with a field guide, a memoir with a manual. The change of register keeps attention fresh and stops a slump from becoming a stop. ReddyBook Club organises titles by mood rather than by bestseller rank, which helps when you have no idea what you want.

Track loosely. A list of finished books is a pleasant artefact and a terrible master. Note what stuck, not how many pages you cleared. Habits built on enjoyment outlast habits built on targets, every time.

Every town needs one room where silence is social. Libraries used to hold that role, and bookshops inherited part of it by accident: people come in to browse and stay because the air is calm.

That calm is fragile. It disappears the moment a shop over-stocks and under-staffs, or fills the front table with novelty tat. The room only works when someone is paying attention to who is in it.

Programming is the other half. A reading on a wet Wednesday, a launch for a local historian, a shelf of staff picks with honest handwriting: these are cheap events with high return, because they give people a reason to arrive rather than to order.

The commercial logic is simple once you stop apologising for it. Browsers become buyers, buyers become members, members bring friends. Play99Win treats discovery as the product and lets the catalogue do the persuading, which is the same bet a good shop makes every morning.

None of this survives without margin. Coffee, cards, special orders, a small paid membership with real perks: the mix matters less than the honesty of it. Charge for the room, and the room keeps existing.

Every town needs one room where silence is social. Libraries used to hold that role, and bookshops inherited part of it by accident: people come in to browse and stay because the air is calm.

That calm is fragile. It disappears the moment a shop over-stocks and under-staffs, or fills the front table with novelty tat. The room only works when someone is paying attention to who is in it.

Programming is the other half. A reading on a wet Wednesday, a launch for a local historian, a shelf of staff picks with honest handwriting: these are cheap events with high return, because they give people a reason to arrive rather than to order.

The commercial logic is simple once you stop apologising for it. Browsers become buyers, buyers become members, members bring friends. Play99Win treats discovery as the product and lets the catalogue do the persuading, which is the same bet a good shop makes every morning.

None of this survives without margin. Coffee, cards, special orders, a small paid membership with real perks: the mix matters less than the honesty of it. Charge for the room, and the room keeps existing.

A second-hand shelf is an argument about time. Someone read this book before you, kept it, and passed it on; the object carries a small record of attention that a shrink-wrapped copy cannot.

Shops that handle used stock learn to price patience. They buy carefully, they cull honestly, and they keep the good spine facing out. The work is unglamorous and it is the reason a room full of books can feel like a conversation rather than a warehouse.

There is an environmental case too, and it is not decoration. Every re-sold copy is a book that did not need pulp, ink, shipping and a remainder bin. Readers who care about that tend to become regulars, and regulars are the only marketing a small shop can afford.

Curation matters more than volume. A shelf that says “here are forty things worth your evening” beats a wall of everything. Reddy Books builds its catalogue around that principle, grouping titles so browsing produces an idea rather than fatigue.

The shops that survive treat the room as the product: chairs, light, a table of new arrivals, staff who will disagree with you kindly. Buy there, and buy used when you can. The habit keeps a street literate.

A small press prints eight hundred copies of something excellent and then discovers that distribution is the hard part. Warehouses want volume, algorithms want reviews, and both want the marketing budget of a company ten times the size.

The physical counter is the fix that remains. A bookseller who has read the thing can put it face-out for three weeks and hand-sell nine copies to the right people. Nine copies is a print run's worth of validation.

That requires a shop willing to take the risk and a reader willing to be persuaded, which is the whole transaction in miniature. Chain stores mostly can't do it: head office buys centrally and the local buyer has no authority over a spine.

Which is why the presses keep a short list of shops that will actually stock them. Play99Win signup carries regional and independent presses as a matter of policy, not as a seasonal gesture, and the difference shows up in the window.

If a small press title has ever landed on you, go and buy the next one at a counter. Pre-orders pay for paper, and paper is where a press begins.

A second-hand copy carries its history in ways a new one cannot. A cracked spine at page ninety, a bus ticket as a bookmark, a name in pencil, a margin note that disagrees.

Some buyers want pristine, and shops keep a shelf of that. But the marked copy is more interesting, because it is proof that the book has already done its job on somebody. It is also cheaper, which means the risk of trying something strange falls to a level where trying is easy.

Shops that handle this stock develop an eye. They know which annotations add and which are just somebody's shopping list, and they price accordingly. The eye is part of the service, like knowing which edition to avoid.

There is a practical virtue in the supply chain too. Re-sold copies keep the money local and keep a book in circulation for a decade rather than a fortnight. Reddy Books new ID leans on that trade, which is why the shelves here look lived in rather than warehoused.

Buy the annotated copy. Read the margin notes. Add one of your own, in pencil, and put it back.

Every reader hits the wall where the shelves are double-stacked and the floor has started to take books. The instinct is a bigger flat. The cheaper answer is a cull, done properly.

Start with the obvious: the ones bought for a course, the ones with a cracked spine and no sentimental claim, the duplicates that arrived as gifts. Box them, put the boxes out of sight for a month, and see what gets remembered.

Then the harder pass. A book you keep because you might be the sort of person who has read it is not a book you own, it is a costume. Keep the ones you have actually returned to, lent and wanted back, or written in.

Sell what has value, donate what hasn't, and keep the local shop in the loop: a good second-hand buyer will tell you which of your discards are worth a tenner. Poker ID login takes boxes by appointment and prices honestly, which is more than the online trade will do for you.

What remains should fit with room to grow. Leave one empty shelf as a rule; it disappears within a year, and the cull earns itself back.

Every small shop runs on a list of odd jobs nobody advertises: rewiring the window light, re-alphabetising a section after a customer left it in ruins, driving an order across town because the buyer is housebound this week.

These tasks look like overhead. They are actually the product. A shop that only sells is a vending machine with a lease; a shop that does odd jobs is a node in the neighbourhood, and nodes get defended.

Staffing follows from that. You want people who notice, not people who can recite the returns policy. Someone who says “the sequel is worse but you'll want it anyway” is worth more than a discount card, because they save the customer an evening.

The same logic applies to the stock decisions. Keeping a local author's pamphlet on the counter, ordering the obscure title three people asked about in a year, holding a reservation until Friday. ReddyBook Club login works the counter the same way, taking the small request seriously.

None of it scales. That is the point. The odd jobs are the moat, and the moat is why the door still opens in the morning.

(Image: “The Vitruvian Anime Woman”, 2022, made with GNU Image Manipulation Program and natalie's avatar)

It is recommended that you read the Virtual Reality: Two Months in post before reading this one, as it might be interesting for you, the reader, to draw differences between those written moments of my life.

After practically a year since I bought my Meta Quest 2 headset, and after all that happened to me with it, I can confidently say that Virtual Reality technology has changed my life, hopefully for the better. I plan to go on some details of what happened between these snapshots of my VR usage, and hopefully some ideas about the future. I purposely left this piece to be written after the Meta Connect 2022 keynote, as I wanted to see the state of the company that I have shackled myself to by buying the headset.

Part 1: Exercise

One big thing that is now a part of my life is exercise. I, as a heavy computer user, had a sedentary lifestyle, and that became worse after the COVID-19 pandemic kicked into high gear and I had to stay at home for protection's sake. I noticed that some parts of my health were starting to go sideways and one of the things I have started doing in the beginning of 2021 (headset bought in Oct 2021) is exercise more, however, it was hard. Attempting to get my legs moving every single day was a very difficult task for a mind that keeps spinning on new things to do, and my hands had to do something as well. I tried running, I couldn't try biking, and so, I was very unstable on the schedule of my exercise.

That changed once I bought the headset and found out Beat Saber is an actual thing. Turns out that when you have already played osu! and know the speed of Camellia songs, translating that into “dancing” was a game changer, because now I could listen to the very high BPM shit that I loved banging my body around to, but now in VR! One drawback of Beat Saber itself for exercise is that it is very wrist heavy and after a while I started developing wrist pain after play sessions, even when adopting a style that was more dynamic. When you move to very high complexity beatmaps, you don't get much of a choice on body movement. Your head stays still, and your arms look like pasta being moved around.

However, I found a different game through Ave. Currently I'm heavily into on Synth Riders and I have a consistent schedule to play it at around 2.5 times a week, with around 30 minutes per session.

Part 2: Socialization and Identity

VR enables a higher level of abstraction around social interaction that I did not think existed before. Let me unpack that.

I would say that a core part of life (as a conscious being) is to play with who you are. As kids, we play with the parts that we align with and refine our image as time goes on, and we add or remove parts of ourselves as things happen. VR takes this to another level by enabling the role play of the self to happen in different physical forms, and you can argue that things like makeup artistry are closely related, and I agree! However, not everyone has a knack for makeup itself, as we do need the skills to move the chemicals around in skin, have to deal with rinsing it out, and in general, that exploration can be slow. When you are in the digital space, that iteration speed is much faster, and this doesn't apply to VR either, text mediums, of which I have communicated in a lot, have the same property.

What VR socialization platforms provide compared to the other Internet communication tools that we have used (voice, video, text) is the body language aspect of it, and when you combine that with the capability of custom avatars, that enables a much richer conversation level. While in a voice chat, I might have synchronization issues with someone else as they might not be fully focused on the conversation, but in VR, there is nothing else to lose focus on (of course, latency, world setting, general environment changes can change, YMMV). In a video chat, there is the constant worry of looking good enough (very complicated when I have my background) and having a constant camera pointed at my face (I continuously have issues with pictures being taken of me by people who do I don't trust to take proper care). I feel more comfortable in a text chat compared to those two, but I miss being able to use the voice modality of myself in that area.

When both parties are in VR and agree on the implicit social rules that would let conversation flow, everything just comes together to create memorable experiences I still enjoy to this day. The tools that flourish from this environment become important value additions to communication, for example: Pens! The QvPen addon for VRChat enables me to simply run wild with drawing whatever I want anywhere and show them to my friends, while not having to worry about gathering paper, finding a pen, drawing things out, probably looking weird, and back and forth. It is because of that richness that I care deeply about how the VR social platform I have most used, VRChat, is currently running. I truly believe that spaces like these can bring massive amounts of life-change to a person, and defending the core tenants that create those changes is paramount to my experiences within VR.

Part 2.1: Possibilities of expression

VRC has implemented Easy Anti-Cheat (EAC), in a move that immediately antagonised the enthusiast core user base of the platform, as everyone there used mods for various things to do in-platform (e.g bugfixes and quality of life improvements). The aftershock of that single action is still being felt to this day, as there is a lot of background to uncover behind that single action.

The platform is basically the wild west of experimentation in social platforms. While things like Rec Room and AltspaceVR are very constrained to the design language of their makers, VRC took a 180 and let everything be made by their users, with minimal oversight. There are extremely good things regarding expression that are brought from this move, however, it is fully incompatible with being a for-profit venture capital funded startup. Startups need to make gigantic sums of profit really really quickly, and VRC has been draining through their funds pretty quickly as well, being a heavy user of cloud infrastructure, hiring a bunch of people, etc. VRC has attempted to gain more profit through a subscription called VRC+ that added you some extra perks (say, more avatar slots), but that is definitely not enough to generate enough revenue (unofficial optimistic estimates put it around 30% of the userbase).

Another avenue of monetization is creating a Creator Economy, of which content creator users, making worlds or avatars, receive direct payment from the users that wish to support or use the avatars themselves. This has been accomplished through an out-of-platform site, the biggest one for VRC paid content (though not limited to just that) being BOOTH, and, being out-of-platform does not enable VRC the company to get a slice out of that. The clear solution being attempting to make a built-in feature themselves for a creator marketplace. However, to make that happen, they must guarantee platform integrity, and EAC is their hammer for that. Believe me when I say that more will come.

VRChat's DMCA process is practically non-existent. The most famous world is a full blown DMCA violation that VRC staff know about. They have posted about this world in their own promotional material, that means that if any company that wishes to destroy VRC akin to the Viacom International Inc. v. YouTube, Inc. case, has pretty good evidence to show in court. Imagine if we get Content ID for 3D meshes.

What would the possible outcomes be?

Look at Horizon Worlds. They're going to play it safe and have announced a Creator Marketplace for avatars coming next year for their own social platform, Horizon Worlds. Of course, you won't be able to get the real 3D mesh of the clothes you buy, modify your avatar to be anything that isn't solid color human, etc.

Part 2.2: Alternative grounds?

When the EAC hammer was thrown, many of the modder userbase moved to competitor platforms, the two main ones that resemble any kind of expression freedom as VRC being ChilloutVR (CVR) and Neos VR. Neither of those have lived up to the massive expectations of being a VRC replacement. However, thanks to VRC's own move (and nebulous future), there definitely is a space for competition to rise up. I did not get first party experiences with CVR and Neos as they do not develop for the Quest 2 standalone platform, I would need a PCVR rig for it, but I'm due to get one someday, for sure. The second-hand statements I have gathered do not paint a good light of those alternatives' respective futures (caused, in general, by bad management), and I would rather have them share their own write ups, and I would be glad to link them up here, if any appear.

And as cited before, AltspaceVR, Rec Room and Horizon Worlds do not apply for the contender of being an expressive platform, and I would say that lives at the core of properly socializing with someone.

Part 3: What to do from where we are

Right now the social platform “metaverse” space is being fought by many, many companies, there are various target audiences, and I believe that none of them will properly “win” per se, but we will have the same plurality of instant messenger apps. Remember when everyone used Skype and moved to Discord? I'm waiting for Discord's major fuck up to happen and have another competitor being funded by VC cash to explode and follow the same cycle.

However, one thing that hasn't died from that IM craze is the core foundations of the Web, as it is its own freedom that enables the competition to exist. I would suggest, then, that a “metaverse” that attempts to displace the current computing platforms of today, built on HTML, JS, and CSS, should follow that same Web model. VR needs its own set of open standards that would let you make your NGINX and your own Firefox equivalents (not direct equivalents, mind you, but a general server engine to serve content, and a general client engine that lets you explore content through a set of interfaces, which at the moment are the keyboard and mouse).

If you can provide that tooling, you (well, the project) would become immortal. Nobody can sue F5, Inc. (current owners of NGINX) or Mozilla for a DMCA violation, they only provide the protocols for content that can come from anywhere else to be viewed, and then the owners of the servers hosting that content would be the liable party, just like the current Internet is right now. There are many possible issues with this idea, but I think it is feasible to exist in the current landscape of VR devices. I do not think Horizon Worlds would need a new Quest 3 headset to even move in that direction, just let creators explore the capabilities of the silicon millions of people have gotten from you (well, Qualcomm, but I digress).

Part 4: Hardware

There were two relatively major hardware releases this year in the standalone headset space. This section is very small as I'm not a hardcore hardware enthusiast.

PICO 4 is what a Quest 3 lookalike would actually appear as, taking the same lessons of the Quest 2, making it a smaller form factor. However, as the headset does not really provide something akin to VRChat, or something compatible with my PCVR friends, I can only consider the headset as a PCVR headset itself, using some sort of link cable or Wi-Fi streaming, which is what I already plan to turn my Quest 2 into already, the upsides being more balanced (fuck the Quest 2 default headstrap), and being on the side of actual competition against Meta's hardware.

The Meta Quest Pro was announced today and it is NOT made for any sort of proper consumer user. It is priced at 1500USD, and it is heavily targeted towards companies that want to buy them in batches and want their employees to “work in the metaverse” and “build their models within VR” (even though they're not properly advertising the platforms that would actually make that happen, so I don't see why a business would want to buy something that is on unstable ground at the moment).

In relatively minor news, the Meta Quest 2 has gotten a price bump from 300USD to 400USD, opening the door for competitive headset makers like PICO to hit that price point. And there are possible leaks of a Meta Quest 3 coming out in 2023, but as it's all leaks and no proper substance, I won't speak much on the matter

Part 5: Conclusion

I hope that the future is bright and that expression can be ultimately defended by those who prioritize that the most. It is a hard battle, but I still have to hope, or else nothing will get done at all. The “neat”ness of VR technology has grown within me in a way that I do not think I could properly handle not having access to it on a weekly basis. You could say this is addiction and escapism, but isn't all of humanity's ventures some sort of escape from our current stage either through fictional stories being played in various ways, or through technology like the Internet which completely changed the grounds on the distances between people?

I still enjoy it myself and have had a lot of fun exploration that can't be properly expressed in my day-to-day activities, but only when I can write this post that I can put it in a longer time frame and draw metaphors that I hope can get across on how much this has become one of my passions.

I plan to write another post on October 2023, containing the next year's outlook on the VR landscape. Until then, good luck on your endeavors. I am open for words on the “blog at l4.pm” email.

Alternative title: “My answer for the next decade of my digital life”.

I am, generally speaking, a data hoarder. I don't really have the money to invest in large ZFS or CephFS or whateverFS clusters of storage, so I try my best with my 1TB drive I use for my entire system (nowadays with a 256GB NVMe drive for the root partition, so another 100GB can be quickly eaten away by datahoard-ness).

If I recall correctly, this strategy of doing things started back in the 2010s where broadband was still shy inside the country (10Mbps down, less than 1Mbps up was common, nowadays in 2022, I'm getting almost 300Mbps down, 40Mbps up, with the same pricing from back then), and a girl needed her electronic music fix. Since my school doesn't have a connection, I started hoarding youtube mp3 files from a random youtube-to-mp3 service, and copying the files to my 2GB phone.

My laptop had way more than 2GB, and there's an infinite amount of music to listen to. What to do? Organizational systems, of course! The first thing I remember doing was ordering folders by increments of 1, where each folder had some 10 to 20 files. So I could grow the library up in my laptop, and copy the most recent tunes to my phone in a very easy manner (delete all folders below 25, for example), while not having to lose the old stuff, just in case. If you know my soulseek, you can see that same system living in 2022.

As time passed, I started gathering different kinds of media, videos, papers, books, and I noticed a pattern: they aren't properly read in sequential order. Say, if I wanted to find some scientific paper I had saved ages ago, and I know what it's about, I would have to either: – Go through each folder sequentially. – Hope that I remember the title.

If you're yelling “booru” out of the top of your mind, then yes, I didn't know at the time, even though I was a heavy user of booru software to search for new images of anime women holding hands, but booru systems are the ideal solution for this kind of problem. Non-hierarchical, tagging systems that have (digitally) lived with us for a decade now, but the ideas behind it come much earlier.

I'm ashamed that it took this long for me to refer to it in writing, but Nayuki's blogpost on non-hierarchical systems is an eye opener to what I've been wanting to have for ages and I didn't know it. It's absurdly long, but it's orchestrated akin to a machine gun of ideas, and I love that kind of stuff.

There are many booru systems out there, but one of them that's close to what I've been wanting is Hydrus. From the website: > The hydrus network client is a desktop application written for Anonymous and other internet enthusiasts with large media collections. It organises your files into an internal database and browses them with tags instead of folders, a little like a booru on your desktop. I have attempted to use it to organize my libraries more than once, since it would “technically” fit like a glove, but I had issues with it, though they are not to say that hydrus is bad, more that it doesn't align with my vision for such a system.

In hydrus, you import your files into the system, you can add/remove tags, share metadata around in the “hydrus network”, etc. But the biggest dealbreaker for me is that once a file is added, its original location becomes meaningless. To be able to refer to that file in the future, you must use hydrus to find it, because everything is in an internal client_files/ directory where the filenames are renamed to their hashes.

Everything still works from a filesystem level, sure, but if you lose access to Hydrus, you now get a client_files/ folder you can't understand anything out of by filepath anymore. That is a design decision that I perfectly understand where it comes from, but it brings me pain (see: the folder organization structure I just showed is not possible to happen inside Hydrus unless I create hacks like symbolic links from client_files/ into the sequential folder structure).

So, if I wanted to make my own non-hierarchical system, it would have to operate as an overlay on top of an existing filesystem, keeping references to the original file paths. That becomes a problem really fast, and that's the main reason why hydrus does what it does by design: renames. If a file is renamed, your reference to it goes away. A system that does not take ownership of the file contents entirely would have to keep track of path renaming.

There are two approaches I have found for this: – FUSE. – syscall tracing.

FUSE is something I have never touched on and there's the possibility of causing high latency on FS operations as things go back and forth from kernel-space to user-space (in theory io_uring could help in this case but I have no idea how it works).

Syscall tracing is possible thanks to the eBPF virtual machine that's in the Linux Kernel. bpftrace is a CLI utility for Linux that is heavily inspired by dtrace which works on the BSDs and illumos systems.

So, well. I guess I made it.

https://github.com/lun-4/awtfdb.

Here it is, thanks to bpftrace integration I am able to make my own “rename tracker” and update the index database file automatically. Because of all of that, I have a much harsher opinion on bpftrace, but that's answered by “it's not 1.0 yet, so don't complain”.

awtfdb is a collection of CLI tools (at the moment, Linux-only, MacOS soon maybe?) that operate on a single index database file, powered mainly by Zig, SQLite, and my power to bother a colleague so hard they make a library. You can include files and their tags into the index with ainclude, search for tags with afind, remove files through arm, etc. Since bpftrace has to be run as root, awtfdb-watcher is provided so that only that dedicated process can have root access, while the others just operate under your normal user.

This goes back to the alternative title. If all goes well, this project is planned to stay around for a long time for me, and is how I want to manage my media libraries in the long far future. It isn't super stable right now, and there's design questions to answer, but I'm hopeful for that future.

It has been almost two months since I received my own VR device: The Oculus (now Meta) Quest 2. A few years ago I used a PlayStation VR at a friend's house, but that was only for a handful of hours, and nothing other than Beat Saber.

I would say it has been one of the greatest investments I've using the money I got from my full-time job. Thanks to economies, I'm not able to just get a powerful setup to have PCVR, which means the Quest 2, with its price range, becomes very enticing (as in, orders of magnitude more enticing). The amount of things and experiments I can do with it justify the cost, and several times a day I've been having even more ideas on what to do with it, it's great.

Because of all of this, I'm now also considering financing my PCVR setup, but that's for the long term (as in, “2-3 years” long term to save enough, if something more pressing doesn't happen by next year).

This blogpost will condense various takes I've had over the months regarding various topics related to VR.

The limitations of Standalone

The Quest 2 is, for all intents and purposes, an Android phone with a big screen, 4 infrared cameras, weird lenses, wild speakers, and buttons! Fucked up! It's absurdly impressive work by the Meta engineers. I believe that the future will be in this standalone space, because you can drive the price down, and cheaper tech means more adoption of that said tech.

Currently (December 2021) a Valve Index costs 1000 USD, then you need a powerful PC to achieve smooth VR, and for me, currency conversion and shipping costs mean that it's too much of a risk for me. The Quest 2, on the other hand, is 300 USD by itself. Sure, import taxes bring that value up a notch, but it's much less risky of a purchase for someone in my situation. And now I just need the “powerful PC” part, since I can use something like Virtual Desktop to play SteamVR games in the Quest 2.

This is not to say the Quest 2 is perfect. It is a phone, and so, performance takes a big punch. If you compare PCVR experiences and their equivalent Quest 2 ports, you can smell the optimization (without that smell box!! It's a fucking smell box).

Sidenote: Privacy

You can remove your connections to Oculus's servers, but in the end, that turns the Quest 2 into a PCVR headset, and while that's a viable plan for me, who wants a PCVR setup in the future, I do not think that is a useful answer to the large amount of Standalone VR players that will begin to appear in the future, see this market forecast, and Steam's hardware survey, of which Quest 2 has the majority with 36.32% (November 2021 Hardware Survey data).

If what we want is to release ourselves from Meta's shackles, I think the only way to do it is to root the Quest 2, and start to develop custom system software. While the former is a hard endeavour, the latter is an even harder one. You would need to: – Reverse engineer the Qualcomm Snapdragon XR2, an SoC which does not have public documentation, and I don't want to talk to Qualcomm to find out if I need an NDA or not. – Write your own SLAM system to counteract Quest Insight – God, the hand tracking data (it's just a neural network, for sure, but oof) – Possibly need to reverse engineer VrApi, that might change as Meta wants things to use OpenXR, but I can't find reliable sources for that one. – Oh god how do we get APKs from the Oculus Store to run on the headset – And a lot more that I don't know of!

There still should be the ability to root so that those reverse engineering efforts can begin. The whole point of the Quest 2 is in the hardware it's giving and some amount of a social platform, but if you jailbreak that hardware, you jailbreak the value of the headset. Enabling freedom while making Meta lose money with the headset (source for the business loss).

Metaverse projects and the point of VR

Right now, there are around 10000000000000 metaverse wannabes, I'll categorize them into 3 areas: – VRChat – Horizon Worlds – The rest (Rec Room, AltspaceVR, etc)

I am of the opinion that VRChat is the only one that will have any kind of foothold in the future because of one singular thing the others don't have: Personal Identity. I have talked about this topic before with many others, but in the other platforms you have a “style picker”-style screen which you have seen in many other games, like The Sims, while on VRC you can have just about any 3D mesh (within the limits, of course). That allows for a much greater array of possibilities compared to the other platforms, and in turn, generally better social interactions. A close friend of mine says that when she's talking to me in Rec Room (when we were trying it out), it felt like talking to a manufactured avatar, but switching back to VRC, she reported it felt like having a conversation with me, Luna, even though I was appearing to her as a completely different being.

I'm not sure if other people would experience this slice of VR differently, and I might have biases on identity exploration due to being transgender and doing such myself, but if someone wants to be a cute anime girl or whatever else they want, they should be encouraged to do so, and being a permutation of a humanoid avatar without legs is not really unconstrained enough to do such experiments with.

VRChat and Geocities

Practically all of the worlds and avatars in VRChat are created by users. The VRChat company provides the client and server infrastructure, and sometimes partnered worlds, for which I don't know the actual creation process of.

You can draw direct parallels from that into how Geocities worked. It was free user content hosting servers, and then someone else came with the web browser, but it was raw, random creativity by other people, of which you had to dig through and find yourself.

Though, Geocities died of unknown reasons. I was too young to have gone overseas in internet space, and the country I'm in didn't really have any sort of Geocities-esque movement, we were diving in by the millions by the time FAANG already became an information behemoth.

One guess I can make, which also applies to many other platforms, is the cost versus revenue math. You are giving people infrastructure for free, and how do you plan to pay for that infrastructure? It isn't even known if YouTube is properly profitable for Google, but AdSense pays for it all. Will something akin to that happen in the VR space?

To take a more positive light, even though I feel a lot of worry in the infrastructure side of things, I don't feel as much in the actual content generation ecosystem. Likely because platforms like Patreon exist, which enables a far easier donation scheme than back in Geocities days. Hell, even Neocities, the spiritual successor to it, is backed up by either donations or a small 5USD/mo plan.

My wanted VR future

Disclaimer: The “My” in that title is on purpose. You might disagree with what I said in here, in any front.

While I don't usually love to think in long-term timelines, I have some wishes here and there.

• VR should be brought to more people, and it's why I can see the bet on standalone VR because it is a great driver for that.
• Metaverse being owned by people, not profit motive. VRChat is not that, as they are a startup and own the entire infrastructure of itself.
  • Can federation solve this? I don't know. I feel disillusioned by it nowadays, you can probably get the benefits by being centralized, but structured like a worker-owned cooperative.
  • I don't think blockchain will save us because of the complexity of the technical hurdles while centralized systems bring development speed and other things, but this could be written in more detail on another blogpost.
• Full-dive VR is not something I directly want in the current state of things, it requires us as a society to think differently about technology, because in virtual, anything is instantly copyable compared to meatspace.
  • There's also the whole “write-based BCI” which brings an avalanche of ethical discussions, which, again, can be developed in a separate blogpost.

Conclusion

I think VR is neat. Its neatness should be defended.

Here's a fun music track: Dom & Roland – Imagination

Hello, if this is your first time reading this, I'm luna, one half of LavaTech and we provide continuous integration (CI) services for the Zig programming language.

We started doing this because Zig could not use their pre-existing CI service for FreeBSD builds, sr.ht, because the compiler became very RAM-intensive and that hits the limits on their CI. We stepped in and we now provide a selfhosted Sourcehut instance where FreeBSD builds can still happen while the self-hosted compiler is not finished yet (Genuinely, I have no idea what will happen once the memory usage is brought down, but for now, I'm happy to provide the service, and wish to continue).

On a sunny day with nothing much to do, I decided to bring better NetBSD support for Zig, and one of the areas I can learn on is the one where I'm operating: CI.

This blogpost outlines the things I had to do to bring NetBSD to Zig CI.

Some background, or, How does CI work

Sourcehut is a “software forge”, which is a collection of tools to manage codebases, either publicly or privately. sr.ht is the main instance of Sourcehut, operated by its original creator, Drew Devault.

builds.sr.ht, the CI service of Sourcehut, works by providing QEMU/KVM virtual machines that run specifically-crafted operating system images, and from them, your project can be built and the resulting artifacts can be tested inside the virtual machine, all via executing SSH into the VM. After all the commands are run without errors, the build has “passed”. An example of builds.sr.ht on the main instance can be seen here

The outline of the available VM images can be seen in the compatibility page of builds.sr.ht, and from there you can see Alpine Linux, Arch Linux, FreeBSD, among others (hell, even 9front!).

NetBSD on sr.ht

Back when I was developing support for it, the compatibility page did not mention NetBSD, and thanks to Michael Forney, it now is!

The history for it is as follows: – The initial versions of the NetBSD image were created by Drew Devault. – However, they weren't really finished. – I began working on finishing it. – Michael Forney also began working on it, in parallel. Their patch has been successfully upstreamed! – Even though we worked in parallel, our changes are mostly equivalent and so, there isn't a need to upstream my changes: – Replacing anita (The automated NetBSD installer) for directly downloading the binary sets and partitioning the virtual disk. This is done to remove the QEMU requirement on the build script, because I am building the image in a NetBSD VM itself, and I was not sure on how to bring nested virtualization to it. – Replacing pkgsrc building source packages with downloading binary packages via pkgin.

Here is a successful test build with the image!

Preparing the CI

Zig CI does not depend on the upstream platforms' LLVM builds. Instead, a self-built LLVM is created via the zig-bootstrap project. With it, you can go from a C compiler to a fully functioning Zig compiler for any architecture. It achieves this via four large compile steps (with the relevant names I will use for them throughout this blogpost): – Compiling LLVM for the host system (“llvm-host”). – Compiling Zig for the host system (“zig-host”). – Compiling LLVM for the target system (“llvm-crosscompiled”). – Compiling Zig for the target system (“zig-crosscompiled”).

With, for example, the zig-bootstrap result for aarch64-linux-musl, you can start an ARM64 Linux CI, where zig-crosscompiled compiles the new commits of Zig. This is possible because Zig is a C compiler as well. Here's the script for Zig Linux CI on Azure of such artifacts working in a CI script ($CACHE_BASENAME).

For NetBSD, the process shouldn't be different (FreeBSD follows it as well), so what we need to do is get zig-bootstrap running on it. This process took a couple of days, with help of andrewrk and washbear, and long arduous moments of the having to cyclically “Start the build, find something to do, mentally switch back to fixing the build”. It all paid off, and I generated fantastic quickfixes in the end! Here they all are!.

The submitted PR is more of a discussion place about what can be done regarding my workarounds, rather than putting all of them into zig-bootstrap (because I am pretty sure some of those fixes would break... every other target, lol).

Spinning up new infrastructure for CI

builds.sr.ht has the idea of build runners for jobs, so you can dynamically add more capacity for CI as needed. Each runner can have N workers, and those are the ones running QEMU VMs.

As we started with FreeBSD, we learned that a single worker requires 16GB RAM (the actual compile requirements are approximately 8GB, but there are a lot of build artifacts, the filesystem for the VM is composed of both the image and a growing temporary ramdisk). And as time passed, we learned that 4 build workers can handle the throughput needed by Zig when it comes to FreeBSD (as of July 2021, this is true).

To have NetBSD support added to the mix, we would have to add 4 build workers to the network, which means finding 64GB of available RAM somewhere in our infrastructure. LavaTech operates in a hybrid cloud model: There are VPS'es in various cloud providers, but most of our services run in our own colocated hardware, shared with general programming. We are able to do so because of Hurricane Electric's free colo deal. Also, Peer with us!

In general, our rack has two power supply units, one big momma computer (of which we kindly hostnamed it “laserjet”, it was close to being “HP-LaserJet-Enterprise-MFP”), and a bunch of blades.

There will not be a hard requirement for this list (or the infrastructure mentioned in this article) to be updated as time passes.

And this is the build runner allocation before NetBSD work: – runner1: Laserjet VM (2 workers) – runner2: Laserjet VM (2 workers)

After talking to our friends at generalprogramming, we were able to allocate two more runners: – runner3: GP Blade (2 workers) – runner4: GP Blade (2 workers)

However, they were unstable as time passed, and we decided to rent a server which would take care of both FreeBSD and NetBSD: – bigrunner: Hetzner Rented Server (8 workers)

runner1 and runner2 were decommissioned to decrease load on Laserjet. runner3 and runner4 were decommissioned because of their instability.

One infrastructure note to keep in mind is that we aren't using all of the available blades in the rack, because of power issues. The PSUs can handle 20A maximum current, and we have both of them so that we can stay redundant. Even though we have 40A theoretical max current to be used, we must only use 20A max, and we were already on that limit. Foreshadowing: A recent failure in the datacenter brought down one of the PSUs and our router was not plugged into both, causing an outage.

A collection of operation problems that happened in the build network while getting NetBSD CI up

Accidental Ramdisk in Production (July 12th 2021)

The issue was first identified by seeing jobs close successfully with “Connection to localhost closed by remote host”. This is bad (it should fail, really), but we knew that it was related to the OOM killer, as we had it before.

We noticed that the RAM usage on affected runners was 50%, even though no CI jobs were running on them. Before deciding to reboot, thinking it was some cursed I/O caching eating RAM (it wasn't, after checking free output), I checked df and found that the rootfs was a tmpfs, not something like ext4.

Turns out I accidentally installed those runners on the alpine livecd ramdisk. I didn't run setup-alpine. The motd as you enter a shell tells you to run it. I don't know how I missed it.

After backing up the important data (build logs), I was able to properly recreate them. Since I knew that more runners were bound to be created in the future, I wrote a pyinfra script to set everything up beforehand, and became a fangirl of it afterwards. Oh well, it happens.

Network Timeouts (July 13th 2021 and beyond)

Sometimes connectivity to the builds service simply times out with curl: (7) Failed to connect to builds.hut.lavatech.top port 443: Host is unreachable.

There are high packet losses, maybe because of one of our network upstreams, but I can't dive deeper into this issue. Hopefully it isn't as frequent.

Docker networking mishaps (July 13th 2021)

Jobs simply didn't start with the following error:

docker: Error response from daemon: Conflict. The container name "/builds_job_unknown_1626160482" is already in use by container "90c8a6c874840755e9b75db5b3d9b223fbc00ba0a540b60f43d78378efc4376a". You have to remove (or rename) that container to be able to reuse that name.

The name of the Docker container running QEMU is decided in the control script, here's the relevant snippet of code:

        --name "builds_job_${BUILD_JOB_ID:-unknown_$(date +"%s")}" \

The script that actually boots the VM containing a CI job is the control script, usually found in /var/lib/images/control (when using the Alpine Linux Sourcehut packages, at least). Sourcehut documentation says that the user running that script should be locked down to only being able to run it so that vulnerabilities on the worker process don't lead to a privilege escalation. That is done via the doas utility, an alternative to sudo.

A missing parameter on the doas.conf file makes the control script run without any environment variables, since BUILD_JOB_ID is missing, it uses date +"%s".

This issue didn't arise until NetBSD builds were merged into Zig. Before that happened, we only spawned a single job per PR: the FreeBSD one. However, after adding NetBSD, we started spawning two jobs per PR. Since they were close in time, the container names would be the same, one of them would work, and the other would crash with the aforementioned error.

VM Settling times out (July 28th 2021)

Some VMs were timing out on their settling, but only when the machine was under load (such as after a merge spree in the Zig project). When running a CI job, a new VM is created running the specific OS image, then a “settling” task is run for it. The task attempts to SSH into the virtual machine, run echo "hello world", and if that works, send the rest of the commands (cloning the repo, running the scripts declared in the YAML manifest, etc).

We were always missing timing on it, so by the time I connected to the runner, it was fine and stable. But by the date mentioned in the subtitle, we were able to catch it in real time, and saw that the load numbers are close to 16, even though the runner has 8 cores total.

Turns out the CI VMs have 2 cores allocated to them, that's hardcoded in the control script.

Current solution: Edit the control script to make the VMs be single-core.

Possible future solution: Buy more servers

What next?

• It works, and zig provides master versions of the compiler for NetBSD!
• It might take a while until I decide to add OpenBSD into the mix.
• All of this should have a status page and log analysis to find CI failures that were caused by operation errors, instead of false-positives due to someone writing incomplete code in a work-in-progress PR.