Web4 Should Run on LaTeX
cynicusrex.com
cynicusrex.com
No one is making you write web apps, websites are still first class citizens for web browsers.
<rant>
Is it not instead the case that today's hipster developers simply want to ignore that platforms exist beside their fancy McBooks? Is it not the case that today's developers prefer a bizarre interpreted untyped language simply because that's what they learned for building websites and they don't want to put up the effort to understand what a type system is? Is it not true that the same developers pull in dependencies reflexivly and never consider long-term maintenance, let alone such weird concepts like supply chain security? And finally did not some hip leaders invent the microservice architecture to cater to these idiots? Because it's nowadays simpler to have a huge, inefficient, pile of services that hides all complexity in the communication between these services than to have one well-designed, efficient system?
</rant>
Sorry for the rant, but sometimes I feel like the industry is actually regressing because many developers are simply lazy morons that hide behind buzzwords.
* is only for those two (and sometimes only one of those) platforms
* the preferred "version" on those platforms in an app, rather than web-based
"Qt for Application Development is dual-licensed under commercial and open source licenses. The commercial Qt license gives you the full rights to create and distribute software on your own terms without any open source license obligations." https://www.qt.io/licensing/
But since you're such a boi genius, make something that can match and get ahead of it and i'm sure everyone will jump on your boat. Until then, your opinion will remain objectively useless.
I say we are the arch phase of the web, we understand the arch and it has served us well and it does save a little bit on materials. What we want to do is create bigger and better engineered structures like the pyramids.
It's been a long time since I used a cross-platform framework so maybe things have gotten better but what used to happen was that I'd run across something in them that didn't work right and then I'd have to do one of:
1. Just not use that in my application,
2. Implement that thing outside the framework which required me to (1) know how to do that thing natively and (2) figure out how to make my implementation work with the rest of the app that was using the framework, or
3. Figure out how to fix the bug in the framework which required me to (1) dive into the framework's internals and (2) probably required me to know how to do the thing natively.
Framework internals are complicated. This complexity is expected--when you are writing a native application you only need the complexity that is inherent in doing whatever it is that application is doing whereas a framework has to be flexible and general enough to serve the needs of a wide variety of applications.
If the framework is going to make me dive into its internals to get it to work there is a good chance that the total effort of making my app run on two or three platforms via the framework will be more than writing separate native apps for those platforms.
Any replacement technology/protocol which does the same "write once, use everywhere" thing and manages to perform better would still need to achieve wide adoption. And I feel like users would still want that to be able to play seamlessly with everything in their lives that they have online.
They've worked pretty closely with the folks who designed java bytecode, to try to learn from their institutional knowledge (repeating what they felt they did well, but learning from their perceived mistakes).
It hopefully will end up really nice, because it could provide us with a machine-agnostic IR that will work everywhere, have high performance, and if it becomes popular on the web, I could it leaking into the desktop space and starting to become a standard there as well.
The web is great because people kept pushing its limits and weren't arbitrarily held back by bureaucrats. Even though Apple and Google have spent untold billions to push the "there's an app for that" line of thinking, the web has survived.
1. Youtube is just a generic video hoster, of which many came before, during, and will come after. As for the whole "pretend Youtube is a thing" thing, I can't remember one actual good video from a "content creator" ever. It's all circle jerking and quick videos on a topic based on least effort.
2. <video> tag has never had a single good implementation. They are all bugged and 100x slower/clunkier than using the built in menu on a CRT TV from the early 90s.
3. Yes, the web would be better if it was just static content. Even without <video> tag. The primary use of a <video> tag is clickbait (actually, almost nobody uses a <video> tag because they don't want to host video, and instead use a youtube embed, so <video> is not even used in practice). For anything else, like movies, porn, etc, it's 1000x more convenient to just download the video and watch it offline in a real video player like MPV. Web browsers still cannot even turn off that stupid "you are now in full screen" banner at the top of the screen, which fortifies my point that <video> is not good for what it is intended for: short one-off clips scattered throughout web pages.
4. The whole idea that it's hard to have a video tag is a product of licesnsing braindamage and the clusterfuck of video encoding standards.
5. The web is not great, and I try to avoid it as much as possible. Using a god damn terminal emulator (the most braindamaged obsolete shit in the world) is 1000x more pleasant than using a web browser.
(Still a bit silly though since the author does mention latex as an alternative too).
UBI is a non-answer to this, implementing it won't allow everyone to quit their jobs and make art instead. Ideally it's meant to be a social safety net, not a way to sponge off of "billionaires paying their taxes" and no ideal survives contact with implementation. We could hope for a marginal improvement on welfare at best (and that would be a good thing, even small improvements count).
The current web can and should be improved upon. Web browsers are in a "worst (that works) is best" situation after having roughly solved the software distribution problem. We need to take these lessons and move on to something better. Seamless execution of software in a sandboxed environment. That's how you solve the current state of the web, build something that is better suited to what people are trying to shoehorn into browsers and they will move all of their bloated apps off of the web leaving only pristine documents behind.
TBF I don't think the author meant to imply that we should "allow everyone to quit their jobs and make art instead". If we agree that the article is largely lamenting the extortive/exploitative nature of modern society cum digital medium than I'd wager what's in mind is UBI as counter-balance to the increasingly extortive/exploitative labor relations many people are living with. We're sorta getting a taste of this with the effect of exceptional unemployment benefits but that's temporary.
HTML lacks numerous features which a first-class document-specification language should have, including grid layouts (only very recently possible), footnotes and sidenotes, and formulae.
The promise of LaTeX is that documents could be defined strictly in semantic terms, though my belief is that this would last for approximately 37.229 minutes after such a proposal actually hit the Web.
The underlying failure of HTML is that it is runtime compiled, not precompiled. This makes things far easier on authors and publishers, and far worse for readers. In the early days, the former mattered in order to incentivise content. Last I checked, incentivising content is no longer the Web's principle problem.
I've also been leaning toward LaTeX as a document speification language for hypertext systems, though I don't believe that alone is sufficient to fix the Web's ills.
HTML has no such requirements. There's absolutely nothing stopping you from posting that in the first place. Throw something at a browser, and it will try its damndest to ingest it --- that's the "runtime compiled" element of the Web.
(It's not a perfect analogy to software compilation, though it can get surprisingly close.)
Essentially: the Web lacks any editor role. That's long been touted as a benefit. At scale, however, and over time ... it becomes problematic.
Moreover, the commercial Web has a selection process and corresponding evolutionary path which has come to be seen as less than optimal for those seeking high-value, high-relevance content. It is in fact hostile to such content, on multiple bases.
Again, LaTeX alone won't fix these problems. It addresses a few issues of structure. The concept of a post-authoring compliation state (ironically itself somewhat present in many web content-management systems, but not oriented around the document or content itself for the most part, but rather branding, advertising, and surveillance instrumenting) is also somewhat attractive but ... problematic. A discovery-mechanism scoring penalty might help (and Google's certainly applied that in other areas).
That aspect has absolutely nothing to do with the technical format of a document.
That said, and despite some quite strong semantic elements to HTML (and HTML5 especially), and the semantic/presentation separation of HTML/CSS, what we're getting, what seems to be encouraged by, and what the adtech-profiting browser vendor and driver of Web browser development seem to be encouraging, is not in fact well-structured, meaningful, high-value, high-relevance documents.
You're absolutely correct that the core of the problem isn't the technical format. But the technical format's become infected by that core problem.
"Whilst that may still pass poorly-structured documents and result in poorly-formed output" - this by the way happens much more often with LaTeX compared to html in my experience.
You seem to be unfamiliar with this aspect of the system?
- The compilation means that there's at least a check for syntactic validity before any old crap is published.
- JS-based HTML can be fully dynamic to the extent that there's no sense of an underlying document at all. There are times when this is useful. That is an exceedingly small minority of the cases in which it is used. The fact that it's often preferable to rerender an HTML document as PDF, simply for readability, let alone archival, should speak volumes.
There is lua-based LaTeX and js-based pdf. Just use html without JS.
"The fact that it's often preferable to rerender an HTML document as PDF, simply for readability, let alone archival, should speak volumes."
The fact that it's always preferable to export a LaTeX document as pdf...
And well, given the amount of people that use overleaf I would say that a lot of people (although writers instead of readers) consume LaTeX.
(Btw, you can compile html too, try printing it as ps/pdf document)
I am not making assumptions about what you do or do not know. I'm telling you how you're being perceived. You have the power to alter that perception. You've failed to use it.
Sure, but I don't see what this has to do with anything.
As for your perception, I don't care :) Keep it to yourself next time please. You too are being perceived in a certain way as well but telling you how would likely be against this site's rules.
TeX itself is a set of typesetting primitives. LaTeX is a set of structurally-semantic, document-oriented macros built on TeX.
Some of those reasons are technical, though as noted elsewhere in this subthread, economic and other factors seem more dominant, and I severely doubt that LaTeX on its own will address the broader scope of issues raised in TFA.
LaTeX is almost never semantic. People are encountered to think only about the document presentation (mostly due to LaTeX's own failures).
As for footnotes being "first class". It's just a macro, nothing first-class about that when compared to HTML's solution.
https://www.pdfa.org/resource/iso-32000-pdf
The LaTeX project announced an a16y project in January of this year. A tool is now available, though success varies.
https://www.latex-project.org/publications/indexbytopic/pdf/
Of footnotes: LaTeX has the macro, it's common across multiple document types. HTML does not.
Various tools and packages that attempted to generate accessible pdfs from LaTeX have existed for aeons, all of they had a common characteristic: they sucked. I am convinced that by this point everyone who cares about accessibility has moved to other formats (like HTML).
"HTML does not"
<aside> is "first class".
I mean, that already failed for LaTeX. To me it's kind of an unwieldy pile of wierd imperative imports. I did write my thesis in it, but for the most part it was "smash these imports together, pray for the best, if it doesn't work, try a different collection of imports, don't bother trying to fix your imports, settle for "least bad" -- which, of course, was better than MS word". I used to do my resumes too, with the same spray and pray feel. Now I just write a markdown file, and anyone that cares that the resume is poorly formatted I probably don't care to work for (many places sanitize resumes to a single format anyways, i suppose for non-discrimination purposes).
Honestly, even though I myself can't use CSS effectively, what little bit I can use feels like CSS is converging to that declarative-layout-that-is-separate-from-document-contents philosophy that LaTeX tried to do, with the major distinguishing point that LaTeX had a huge "start with sane defaults" selling point, and CSS most definitely does not start with sane defaults.
I guess the point I was trying to make is that if you just type only non-layout LaTeX it looks reasonably good, yeah, it does format it with super-wide margins (shrug), but if you do that with html you get mouth spiders.
https://codepen.io/dredmorbius/full/KpMqqB
There are a few bits of that I'd fiddle with, but the idea of defining margins and avoiding text that runs flush to gutters, as well as defining most of the whitespace in rem & em units rather than pixels or points, is a big part of it.
Another alternative would be a stock set of document types for which users could choose from predefined styles. The document would specify the doctype, the user the styling. That could include various branding or simiilar components, but they'd be reasonably minimal.
The ever-encroaching headers, footers, sidebars, floating interstitials, etc., of the modern web really need to go DIAF.
You need to try TeXmacs (which, despite its name, is not based on TeX ;-) ). www.texmacs.org
If layouting (an computationally expensive step) is done at "run-time" what would precompiled even mean?
But HTML is good enough for people to keep using. And LaTeX is barely good enough after four decades for people to keep using.
My BSc thesis wasn't the first thing I wrote in it, but it likely will be the last one. The idea is good, the execution is a resounding meh.
Hopefully, whatever they do can be intergraded to offline latex tools like latexmk.
I still really disagree with the poster though. Latex is not a good fit for Web pages.
Quip has extremely basic editing features, to the extent that I sometimes find it stifling. However, it does a great job at providing 95% of what most business documents need: several headings sizes, paragraphs, numbered and bulleted lists, and the ability to embed pictures - with great collaboration tools.
Quip as a medium is like Hacker News comments. You don’t have a lot of formatting options to work with so you focus on the content rather than messing about with styling.
Most business documents are written and read in a short period of time. Both of the tools show you who from your organization is reading the document for collaborative reviews and allow people to add comments on content inline. Google Documents allows people to make suggested changes that the author/editor can review and accept, incorporating the edit into the document; or the author can share the document and allow people to make changes to it directly.
For example, when my team is having planning meetings, or we are reviewing a project plan, etc. the primary author will often project/screenshare the document, while everyone else also loads it on their computer. We can each see where everyone else’s cursor is (or highlights) and all edit the document simultaneously (if the author wishes) or leave feedback/suggest edits - that we can all see.
These kinds of features matter more in a business environment in my experience more than the ability to format documents in complex ways.
Personally, I find Quip too simplistic, because it does not handle things well like having multiple paragraphs plus a code block in a numbered list item. Google Documents can also have issues with things like this, but I rarely run into something crucial that I cannot do. (But it does have missing features: for example there is no way to add line numbers to a document — but these are less important now that the convention is for everyone to review the document on their computer simultaneously, rather than printing them out).
I find Word to be the most powerful of all of these editing tools and have the easiest time getting it to do what I want. However, (at least the versions I’ve used) seems geared around writing and saving documents locally. It would be my choice of tool if I had to write a long business document and Google Docs wasn’t fitting the bill.
There’s probably a way to set up collaboration features with Word like with the other tools these days, but the “best” collaboration I’ve seen has been through SharePoint which was painful: people had to “check out” the document in order to make changes, etc. I imagine that with Office 365 Microsoft has something better now but if they do I have not had a chance to use it.
Quip and Google Docs “just work”. They are web applications so there is no difference in what is supported between OS versions like with the Word.
In my career as a software engineer & businessperson I’ve rarely needed more than these types of basic text editing tools to collaborate with and convey ideas to my colleagues. Making collaboration simple, including the ability to simultaneously edit a document, or enabling people to read a document at their leisure (asynchronously), and add comments/suggestions/edits - which always refer to the authoritative latest copy (none of this monkey business with emailing around copies of Word documents) provides far more value than advanced editing features would.
If a person can’t get their point across easily using Word’s defaults, perhaps customized a bit by choosing their preferred font, including diagrams where necessary, then I’d question whether the difficulty is the editing tool or something else.
Unless you are producing specialized documents such as academic research intended for publication, or legal documents intended for submission to a court, etc., in my experience business documents rarely need more formatting than Markdown can produce; and easy real-time collaboration is a massive value add.
Web 2.0 added more interaction, more tracking, and more advertizing.
Web 3.0 adds crypto scams.
If Web 4.0 is going anywhere near LaTeX, it needs to add not just the typesetting language, but also the whole system of peer-reviewed academic publishing, and you have to publish at least 4 websites a year or perish.
Talk to Elsevier. I'm sure they'd like another source of cashflow.
This was either written by a 20somethings who wasn’t there in the “good times” he is describing, or by someone who was there but still has the mindset of someone in their 20s.
I don’t know which is worse.
https://www.wolframalpha.com/input/?i=4+trillion+dollars+div...
The other direction, constraint-based layout, is more useful. A is above B, C is to the right of D. Put all those into a constraint analyzer and come up with a layout. This is a 2D geometry problem and should be solved graphically, not procedurally. Nobody knew how to write constraint solvers thirty years ago, and they can be compute-intensive, but those are solved problems now.
A good example of a system which does this is the 2D sketch system in Autodesk Inventor and Fusion 360. Those are true parametric constraint solving systems. Plus, unlike web systems, they can do curves.
go latex! Make everyone miserable, not just the math people.
I think his frustration is misplaced on the medium of content. He is addressing the same issue everyone else has, the fragmentation of information. The internet isn't a one stop shop anymore. You are forced to cycle through huge amounts of reposts and micro-posts to get anything useful.
Even if you're looking at mobile displays, there's a finite set, and those can either be punted or targeted precisely at the 4", 5", 6", 8", and 10" device sizes which are most common.
Using an e-ink device, I'd really prefer more Web content were paginated. I tend to rely on the EinkBro web browser for anything over a page or two in length simply because I can page through that rather than deal with the fussiness of scrolling.
Once you reach a sufficiently large display, physical pragmatics come to the fore. We've settled on books that are mostly 8--14" or so diagonal measure because that's what's convenient to hold and read. There are smaller and larger exceptions (pocketbooks and other miniatures, coffee-table books and atlases), but those are 1) less convenient and 2) clear compromises between format, material, ease of reading, and other factors.
For a sufficiently-large desktop display, a standard quarto/folio print format displayed 2-up is usually preferable. Scaling from that or presenting 1-up is really the desktop / window manager's problem.
So, at the end you are going to serve html at the browser.
You can do that with web1
Both sides of that fail.
You're now shifting the goalposts.
Yes, ePub is based on HTML. It is also contained, standardised, and structured (not dissimilarly to how LaTeX itself is based on Tex but with standards and structure). It's also not the only fluid document standard, though the other I'm significanlty aware of (mobi in particular) are effectively proprietary.
ePub exists and is good enough.
It has some issues of its own, including an overly-strong foundation in HTML which could well lead it to many of the same issues plagueing the Web. In practice, to date, it's largely avoided those pitfalls.
But again, that's really not what the question I was answering was about. Rather it was in targeting multiple sizes of devices. And I think I somewhat addressed that.
LaTeX-produced documents can be extensively styled, that's handled via a stylesheet.
Rappaport's Philosophy of Computer Science is one book I'm aware of created via LaTeX. It's quite readable:
http://www.cse.buffalo.edu/~rapaport/Papers/phics.pdf
I'm trying to remember which book it was that I'd encountered published as LaTeX source which I had to comile myself. I can only say that it was quite elegant.
There are numerous alternative templates for books available elsewhere, e.g.:
If your standard is to just throw information with random formatting on a screen that ultimately cannot be read, then yes latex can do that for you.
But if that is your goal, pure html suffices.
It's a good title because I laughed when I read it and it made me want to read the article.
I believe this is just an elaborate complaint: The author simply expresses the disgust that many feel of how the utopic promise of the internet and digital technology more general (as captured, say, in the early TED sessions) has translated into a dystopia that does not seem able to find a bottom.
Maybe the promise was fake to start with, maybe mixing a pliable and ultimately "social" technology with a broken economic/political system could never have produced healthy offspring.
In any case, we are in a sort of purgatory, waiting for some miracle to absolve people and make them feel good again about their trade.
Seriously, LaTeX is an abhorrent language, mixes paradigms even though they shouldn't be. I agree that content should prevail over style, and that's *precisely* what LaTeX doesn't do. It's a language devised for nice typesetting (of maths mostly), with paper in mind as this was the most common information medium used in the 70s, and not fit at all for modern display techniques, with viewports of different sizes and characteristics.
What HTML, CSS and Javascript got right is the division of content, form and function. Even though modern trends tend to abuse this to produce bad websites and we should make more simple designs, there is no way LaTeX will acheive this at any point.
However it focuses almost entirely on the idea of documents/information retrieval.
The web would not be at today's scale if it focused on written-word documentation. For better or worse, most individuals are spending their time online for entertainment.
But the idea that we could make an internet that separates concerns again is very tantalizing to me. What if web services served data in a standardized, presentation-independent format, e.g. articles as plain html? What if those services then offered web apps to view content and competes to make the best viewers? Could we get back to a content-first model that still encouraged business to flourish?
The web is no longer about communicating, especially the parts the author complains about (framework-driven apps, JavaScript, social networks, etc). It's about working, creating, and broadcasting to large groups. That's not necessarily a bad thing (although if you write a blog no doubt it feels a bit bad having to fight for attention with tweets and stuff).
The very simple fact is that the web has changed. It's not a place where people publish text articles for their audience to read any more. It's a distribution platform for users to create things with rich, diverse apps. There are a ton of problems with it (privacy, security, ownership, sharing, etc), but none of them will be fixed by moving to a different technology. The problems the web has are people problems, not tech problems.
Would be interesting to imagine a parallel universe where the web is a bunch of Markdown documents linking to each other. Like Obsidian, but global and shared.
My version of "we" wants:
a) Plain websites that are just documents with links and navigation.
b) Yes browser based applications please.
c) Yes native apps please as the browser is severely restricted (and for good reasons) on what it allows those apps to do.
Each part is equally important to "we"
It has not. Many native applications are doing just fine. I have one native desktop based as a product with about 40,000 customers and is growing.
The fact that 90% are doing A does not mean that the rest 10% doing B are unprofitable or less important. "Niche" markets (I could hardly call native apps niche yet) can be very attractive.
https://github.com/hughjonesd/huxtable/blob/master/R/latex.R
TeX is an abomination from hell.
But opening an interactive prototype in Figma just by clicking a link or planning an event with Doodle just by clicking a link is not something I’d want to miss out on just to remove ads and cookie banners from websites. Sure loading speed would also be increased, but internet connections are getting faster and faster and most western countries have quite good coverage of fast networks.
The author also states that he/she doesn’t want to learn new web tech all the time, while server rendered PHP still works and powers a large portion of the internet. The web platform also doesn’t really have breaking changes, so your past knowledge still applies and old pages continue to work.
To me this is just an overreaction that doesn’t consider lots of usecases for the web.
That's easier said than done though, with everyone so invested in this modern game of "emperor's new clothes" that is the modern web.
Today i finished an improved version of that converter in AWK, intending to use it to generate pretty PDFs from markdown, but aside from that, i wish not to interact much with LaTeX itself anymore..
Maybe AWK is an ugly language, but its an on-board tool on all systems im using. In terms of getting it to run its superior to Haskell.
A better plan would probably be to come up with some markup language for the internet that just describes the content and appearance of websites, rather than having everybody download little programs to generate websites on the fly. But of course that is the sort of pie-in-the-sky dream that we can only hope will occur in the far future of 30 years ago.
The problem with HTML is the dom and all the clutter around it. It's not really readable, and require browsers to take decisions for ambiguous markup, which involves bloat.
I generally agree with this - many websites are insanely over complicated. I should need to download react/angular/<popular JS framework>, load some JSON files, download a 100 MB title background image - just to read some text.
That said, I also except that _some_ websites do require these capabilities. But I believe they should need to justify it. If you can get away with just HTML/CSS, then you really should.
The way forwards would be for search engines to lower the ranking of websites that are bloated for no real reason. But this of course will not work when Google is the major search engine - because their entire business model is based on injecting some JS into a page and tracking you. They even make money up-selling you on their cloud services for resource intensive content.
Given that, the way forwards is actually likely to come from a browser supported advertising model - that respects anonymous browsing but also provides some guarantees to the advertiser. Then, size, bandwidth, position, etc, could all be factored in. The browser could even figure out whether the ad is rendered in a place where the user could have seen it.
> Reddit should copy Hacker News
Reddit should revert to old Reddit and refresh the UI. I'm sure they had their reasons, but their current website is actively hostile towards mobile users.
> billionaires should pay their goddamn taxes so we can implement Universal Basic Income
There aren't enough billionaires for everybody to stop working and it's a slippery slope. I know multiple generations of people that live on government handouts rather than work - they are not motivated to work at all. Whether or not on purpose, it encourage the creation of a surf class.
I do except that something needs to be done to address increasing wealth inequality, but it should be a self-sustaining system.
> Billionaires should pay their taxes so we can finance giving everyone basic income, creators won't have to monetize the web, and humanity can live in peace.
Ah, what remarkably wishful naivete. There are so many faulty assumptions in that premise.
It assumes billionaires paying taxes would finance universal basic income. It would not.
It assumes there will always be billionaires to tax. But if we remove competitive incentives, there may well be few or no billionaires. (It's a testament to capitalism that it has not only raised myriads out of poverty, but also created legions of super wealthy.)
It assumes people who receive universal basic income will "no longer be pitted against each other." But competition will always exist; it's intrinsic to nature itself.
The author assumes people will be more virtuous if only we didn't have to work for a living. This is not true. Humans are terrible, and most of us are not bright creative luminaries who'd spend our time in productive or creative pursuits. Rather, most of us are lazy and selfish by nature at best, self-destructive at worst.
It assumes UBI would create peace for humanity. This is not true. Fighting, competition, and resource claiming is intrinsic to nature.
In short, this article has a clickbait headline -- it really has little to do with LaTeX -- and a remarkably naive conclusion.
> Rather, most of us are lazy and selfish by nature at best, self-destructive at worst.
"Speak for yourself."
Pagination is for paper, not screens. Thus, PDFs, LaTex, and companions, have no place on the Web. [I wrote about this recently here.](https://jonreeve.com/2021/05/stop-making-pdfs/)
As a long time troll myself, I think the whe LaTeX is just a vehicle.
Sadly, nobody else was interested, and they probably still won't be. One difference? in 2021, I agree with them.
The problem is simple: client/server model
The solution then is to migrate towards an alternative.
You can provide a plaintext experience if you think it's better, but clearly you don't.