Guido van Rossum: “PDF Must Die.” (2014) | Hacker News Reader

Guido van Rossum: “PDF Must Die.” (2014)

twitter.com

31 pointsshubhamjain60 comments

PDF was made for printing, not displaying. As someone who has been in the printing business for a long time, PDF was a god send, no more worrying about missing fonts, no more page margin differences, no more driver discrepancies.

What you see is what you get, wherever you open it, wherever you print it, that is huge.

Emphasis in "printing, not displaying".

The objection should not be "PDF must die", but "people mustn't use PDF for other purpose than printing".

> What you see is what you get, wherever you open it, wherever you print it, that is huge.

Not really, it depends on the PDF software. I have printed the same bill with SumatraPDF, Acrobat Reader X and Acrobat Reader DC: different fonts, different sizes and missing logos...

That's why when you're trying to print professionally, you're gonna use the original software and not something that try to re implement it. I know it's sad, because open source, but when a (good) standard exist that is closed source, we have to made due, till something arguably better exist.

If the underlying document doesn't embed the fonts it uses, then the reader would perform substitution. This causes different outputs on different readers.

It's more of a PDF producing application problem that fail to either embed the rather uncommon or proprietary fonts.

Usually it is the PDF that is at fault, or rather, the assumptions made by the author (or the authoring software). A PDF meant to be portable and usable for a long time should embed all fonts used. If you get different fonts and sizes in different viewers, that tends to hint at a lack of embedded fonts, leaving the PDF viewer to substitute using its fall back fonts.

Most PDFs I come across tend to come with embedded fonts and work just fine on any decent PDF-viewer.

portrait page layout on a landscape screen when what we really want is multi-column landscape layout

PDF1/A is a great standard for printable documents. Because it doesn't try to do anything else.

Regular PDF with embedded flash, quicktime, HTML/XML forms and/or javascript, is a nightmare.

A few years back, I had to borrow someone's computer to fill out the UK passport form. It was full of extensions that only worked in the latest version of Adobe Acrobat - which wasn't available on Ubuntu (at the time). I couldn't even print the form and fill it by hand.

Yeah, it’s bad when governments depend on proprietary formats like that. It would be better if they would use a simple web app for entering the data, and then generate a PDF server-side. (Or even better perhaps, just let the whole application process live online.)

Doing forms in PDF directly is not the most unreasonable idea.

The problem is it's not one, but three ridiculously overcomplex specification: https://en.wikipedia.org/wiki/Portable_Document_Format#Inter...

It's fine if it's the easiest method available with universal support. In this case in failed on both counts.

He didnt add these important words: "on the web"

PDF as a primary info exchange format online is stupid and needs to die

All these people pointing out that PDF was for printing, not for screen display, are I believe (at least partly) incorrect. I distinctly remember PDF being touted as a screen display technology.

Adobe (and the world) already had a printable format: Postscript. But PostScript was interpreted and had loops and other programming constructs that made it hard to stream and to add hyperlinks to. PDF was explained to us as 'unrolled Postscript'. PDF was also claimed to be much faster to display as it didn't require a interpreter although NeXT (and SGi) used 'Display Postscript' for their windowing systems.

As someone who was involved in print media at the time the font & raster-image embedding was also a godsend but that was a characteristic of the PDF file format rather than the PDF drawing technology.

PDF was deemed so suitable as a display technology that the Mac OS X window server was hailed as 'Display PDF' on launch and a big deal was made of how a Mac could not only display exactly what a printer would show but that high-resolution printing was already built in: everything was essentially already being printed to the screen. Although seldom mentioned theses days I assume the Quartz 2D render is still Display PDF?

https://en.wikipedia.org/wiki/Display_PostScript

He doesn't like PDF because it doesn't reflow to fit his screen, but that is precisely the point of PDF.

I agree with the comments pointing out that PDF was not primarily intended for screen usage. However, it’s quite a good format for typeset text. What if the problem is not really with the format itself, but the way it us used? What if the PDF file was customized to the specific device they will be opened on?

It would probably not be impossible to take EPUB source and then convert to PDF via (La)TeX, with good typograhic settings for the specific device. For my laptop screen the page would be around 28 centimers wide, perhaps set in three or four columns. For my smartphone, one column is enough. With this setup one would not have to scroll up and down while reading the columns, which is the case now when reading e.g. scientific papers. And because of the great quality of the TeX engine it might be easier to achieve a good layout than hacking around with HTML/CSS/Javascript.

I don't have any problem with pdf's, and I think they are wonderful, specially for reading books. They almost simulate same experience of reading old fashion book, but in computer ! I like the idea of not being responsive. Because let be honest being responsive would come with so much headache. Every device then render it in a different way.

But I think the problem is software's we use as pdf viewers. Companies tend to ignore pdf's. They think if they can render pdf's , that is enough.No it isn't. Some pdf viewer cant get fit-to-width right.It kind of feels like software companies tend to ignore pdf's.

Apple has done wonderful job with their word lookup functionality . If English is your mother tongue then you don't have fucking idea how helpful is that.

I am PC user , Ask Cortana functionality (in pdf's) in Edge made me ditch Google-Chrome for Edge entirely, and believe me I spend all day in browser 24/7 and my whole life was in chrome before ditching it. (just imagine how important that simple functionality is for non-English speakers) I don't know how to emphasize it more, but believe me, that simple Lookup (without opening new browser tab and searching for "define $word" is the most functionality most of non-English users want).

p.s. I wish google add pdf word lookup functionality to this : https://chrome.google.com/webstore/detail/google-dictionary-...

From an additional tweet, we can see that Guido's biggest reason for disliking PDF is the text not being responsive.

This is actually my FAVORITE thing about PDFs. They're like pages. They perfectly replicate the book medium they are often taken from. If I want to make something bigger or smaller, I can do that, and it just works with zoom. It's not trying to guess what I'm trying to do and adjust things in ways I don't want.

I'm sure it's an unpopular opinion, but I love the PDF.

They're delightful to read on mobile too with a reader that maintains zoom and allows continuous scrolling. I don't mind flipping the page but it's far more friction than flipping the page of an actual book since you do it about 5 to 10 times more.

What mobile PDF readers do you recommend? I’ve been longing for zoom fixation and continuous scrolling but had no idea this existed in current apps.

Foxit PDF Reader does this beautifully. It's far better on iOS but still works better than the alternatives on Android.

Trust me, it is a DREAM to read PDFs on mobile with Foxit. I avoided PDFs for years because iBooks and Kindle suck on mobile. Acrobat failed to even render the text in a PDF I tried so that was a non starter.

Cool! Will try Foxit then.

I just wish that PDFs were more developer friendly, from what I've seen they're kind of a painful format (not that the modern web page is any easier to read the source of but still)

PDF is a lot easier to handle if you consider it binary output, just like PNG or JPEG. For a lot of developers it makes more sense to use one of the PDF emitting tool chains or libraries available instead of working with the PDF file format directly — the most well-known option being LaTeX of course.

I have no issues with generating PDFs, I think it's great for that. It's the fact that it's a format that's almost impossible, or takes an extreme amount of effort, to retrieve information from once it's been rendered that I don't like about it. It's more akin to an image format than a text format.

I thought I hated PDFs until I got my first Macbook and discovered Preview and two-finger zoom. Turns out, I just hated Acrobat and zooming with buttons.

My gripe with PDF is that it's the standard format for academic publishing, rendering a whole mass of scientific knowledge largely inaccessible for text processing purposes. I've wanted to analyze the Libgen archive of journal articles for a long time but have never found an adequate solution for extracting text from PDFs. Any suggestions on this?

Pdf2txt wasn't helpful?

http://manpages.ubuntu.com/manpages/precise/man1/pdf2txt.1.h...

Sure, the Linux tool "pdftotext" works just fine for this. Two small caveats: ligatures get converted to proper Unicode ligatures and not their ASCII fallback (as one might want or expect) and of course complex mathematical formulas are rendered badly.

I've tried both pdftotext and pdf2txt and I remember not being satisfied with either. Neither seem to handle non-ASCII characters very well, but I'll take another look soon though.

https://pdfbox.apache.org/

PDFs are great for reading, but nothing else. If you want to do something with the text, then plain text or some derived format is better. I would include searching as "doing something". PDF viewers can search, but it's always more painful than searching plain text, and I would say if you're using search to simply locate your subject in the file, then the file author has failed to provide a sufficient TOC/index.

To me, the case in point is the PDF documentation of LaTeX packages. The PDF files are great as a demonstration of what the package does, but they are awful for learning how to do it in your own documents. I am writing LaTeX in plain text, in an editor. I want the package documentation in plain text, in my editor. I sure as hell do not want to fire up a separate PDF viewer and then start flicking through pages, just to discover the correct package option incantations.

I love PDFs, but I only love them as a prose reading experience: open the file, start at the top, read until the end. When I make PDFs for other people to consume, I usually make three different layouts: print, screen, and tablet/mobile. Thanks to LaTeX, that's easy to do.

PDFs are great. Things look the same everywhere. Word docs on the other hand god help you if there is some minor difference in Word version / printer model etc and you're shooting for consistency.

Quoting https://twitter.com/gvanrossum/status/533337876415533056

  @eugeneglybin Proprietary though it is, DOC[X] is easier to reflow to fit my screen. Also, I guess PDF fans don't use mobile screens enough.

Reflow isn't about PDF per se, it is about the underlying reader being used.

That said, it is certainly a much better alternative than DOC[X]. Different versions of Microsoft Word tend to produce different output for the same document.

With PDF, you are guaranteed the same output upon printing to any device; which isn't quite the case with DOC[X].

Are there really any PDF viewers that does reflow? I’ve never heard of that before.

The PDF creating application needs to ensure that the document is accessible. Adobe Reader (on desktop) does support reflow [0]. Adobe Reader on Android also supports reflow [1].

[0] https://helpx.adobe.com/acrobat/using/reading-pdfs-reflow-ac...

[1] https://www.quora.com/Whats-the-best-free-PDF-reader-on-Andr...

I use this tool to reflow PDF papers for an e-book reader: http://willus.com/k2pdfopt/

Is there a container format for HTML that includes assets such as images used in the document? I think having a single file is important. Firefox (possibly others) allow you to save a website, but you end up with a folder full of assets which isn't very user-friendly.

Replacing everything with data-uris could be an option, but this would have to be some sort of standard for documents usually sent as PDF.

Websites could then create the "raw" document and display it in an iframe, so the website's template is separated from the document.

http://stackoverflow.com/questions/2429934/is-it-possible-to...

I am wondering why there is still no compiled HTML format in any age when web pages can render almost everything (at least, I can't think of any from the top of my head). With base64, images can be embedded without any external dependency and I think it should be easy to reduce what Javascript can do to address privacy concerns (like, disabling XHR).

A web page renderer is more ubiquitous than PDF. Web pages are responsive and everyone can build one with just a simple text editor. So why do we need PDF?

MHTML?

> "A web page renderer is more ubiquitous than PDF"

Just any browser can render a PDF as well. PDF is good because it guarantees the formatting intended by the author.

Most mobile browsers can't the, last time I checked. Although, it's easy enough to read PDF with so many options on Android, but opening another viewer and then zooming, scrolling is just painful. PDF might guarantee the formatting but I think it should only matter when it comes to printing.

Besides that, fixed size containers and fluid grids give plenty of flexibility with web documents. So in case, the author intends to have fixed formatting, I think it's achievable with HTML and CSS.

And PDF is bad because it guarantees the formatting intended by the author. The author's formatting might work on a single screen, but it sure as hell won't work on the multitude of devices that access the web.

PDF guarantees the end print result, but it shouldn't be used for anything else.

Well letter, legal vs A4 is a concern. Some printing drivers tends to alter content in the «fitting» process making some pdf hard to read. And screen tends to be in weired and diverse height vs width ratio.

US standards are ridiculous. At least A1/A2/A3... are homothetic.

EPUB? It's zipped rather than compiled but that's an implementation detail when all platforms can unzip in memory.

So lets see, you would prefer (say) schools to upload letters to parents on their Web site in the original docx format?

Because that's what will happen if you kill PDF.

+1, pdf cant do this http://explorableexplanations.com/

Probably it's actually the other way around.

PDFs usually fulfill lots of typographic criteria for better legibility like use of justified text and ligatures that are hardly used in HTML or badly supported.

The only real problems are with two column text moving around is annoying and you often have to zoom in. However with a modern touchpad offering 2D scrolling and pinch to zoom this is hardly a problem anymore.

”with a modern touchpad offering 2D scrolling and pinch to zoom this is hardly a problem anymore”

I really can’t agree with you there. Despite that, the reading experience suffers a lot IMO.

Yeah maybe I was a bit too enthusiastic, but it's a dramatic improvement compared to before.

I agree :)

Meh. Let's be real, PDF has its use, it's all a matter of who uses it how. That's exactly the same with HTML. Which of course is something you can't really use for digital contracts, for example.

Wouldn't it be better to educate users to choose the right software tool for the right job, instead of generalizing the problem with shortsighted assertions?

Smart guy, stupid tweet. If you decide to take an extreme position to attract attention or vent, at least say something intelligent.

I don't like PDFs but you can sign them with certificates. This makes them great for busness contracts.

You can sign any lump of plain text with a GPG signature. I know I'm dreaming here, but...

Honest question: How much traction has PDF-plus-certificate gained versus plaint-text-plus-signature?

Lots when you consider that almost everyone on every platform has a PDF reader to work with that signature, whereas not many people have GPG or GPG tools that are friendly to the non-technical users.

https://www.adobe.com/uk/epaper/tips/acr5reflow/

https://github.com/mozilla/pdf.js/issues/4816

Both PDF and HTML are inferior substitutes for proper TeX formats.

What do you mean? TeX is a source format, not a display format.

Reply on news.ycombinator.com