@AlanMunn -- Less time? Last year it took me only about 10 hours to do our taxes. This year it was closer to 35. The government website had so many errors that it took me quite a while to find the schedules I needed. And the fill-in forms can be printed out with filled in data only with Adobe Reader; with any other pdf reader, they print out blank. (The government used to be prohibited from specifying any specific manufacturer's product.) Not an improvement.

@JosephWright Hello Joseph! I noticed some characters (namely < and >, the ones I noticed) are not rendering correctly in your blog (< appears as < and > as >). See this page about .dtx files: texdev.net/2009/10/06/a-model-dtx-file

@barbarabeeton I use Turbo Tax, so no government forms for me. And the reason it was easier was the fact that they've removed all the deductions for unreimbursed employee expenses, which was the bulk of the work for preparation in prior years.

@barbarabeeton Used to be? Did something change?

I was wondering if anyone could suggest an alternative to the standard CC for letters, which might use more space. I'm trying to fit a letter to two sides, and the CC at the end pushes it over the edge. I'm guessing it's in a box or something.

Good bye, Notre-Dame :(

@yo' :'(

@PhelypeOleinik Thanks: the Markdown conversion left a few mistakes

@DavidCarlisle, @UlrikeFischer I've got breakable hyperlinks working in dvips, but there is a wrinkle which really needs a PostScript expert on it

@UlrikeFischer, @DavidCarlisle I might drop the author of hypdvips a mail

@UlrikeFischer, @DavidCarlisle The problem is that object definitions don't work: they break the PostScript. But only if the link gets split. I'm sure it's solvable: probably need to move the object ref. out of the dictionary.

@UlrikeFischer I have a feeling the user/goto split is a waste of time: you end up with all the stuff in the same dictionary, and other than pdfmode it's all just run together

@UlrikeFischer, @DavidCarlisle Actually, I have a plan ;)

@JosephWright I will try to look at it. What about the non-grabbing links. Do you need the width or the height of the link?

@UlrikeFischer I'll have to look at that yet: almost done with breakable links (I hope!)

@yo' :'(

@yo' :(((

I had a great idea, making myself a moderator by adding ♦ to my profile name, but unfortunately the profile names can't contain the character :(

@PauloCereda I'm excited.

@Skillmon <3

@Skillmon aren't there many very similarly looking unicode characters? @DavidCarlisle could help :-) Also please note that we used to have a person here who renamed themselves to "Moderator", and after some discussion, they were asked to change the name as it was confusing people around.

@yo' the possible character set seems very restricted, only alphanumeric plus a few others. I just wanted to try whether it's possible and had of course no intention to actually keep it.

@Skillmon note that you can't change your name twice in a row :D If you messed up, contact Joseph Wright, AFAIK he could help in that case

@yo' I didn't change it because the symbol was blocked. Didn't try anything else.

Everybody, please don't look at my profile. I currently have 999 profile views, beautiful number!

@Skillmon ooh

@yo' ⌑ ⧫ ⬧ ⬪ 🞝 🞟 ◆ ♦

@DavidCarlisle ooh

  U+2666 BLACK DIAMOND SUIT     ♦ ♦ \vardiamondsuit \ding{169}
  U+0020 SPACE     \space
  U+2311 SQUARE LOZENGE     \sqlozenge
  U+0020 SPACE     \space
  U+29eb BLACK LOZENGE     ⧫ ⧫ \mdlgblklozenge \blacklozenge
  U+0020 SPACE     \space
  U+2b27 BLACK MEDIUM LOZENGE     \mdblklozenge
  U+0020 SPACE     \space
  U+2b2a BLACK SMALL LOZENGE     \smblklozenge
  U+0020 SPACE     \space
 U+1f79d BLACK TINY LOZENGE
  U+0020 SPACE     \space
 U+1f79f BLACK MEDIUM SMALL LOZENGE

@DavidCarlisle 🅄🄽🄸🄲🄾🄳🄴 🅂🄴🄴🄼🅂 🅃🄾 🄱🄴 🄰 🄽🄸🄲🄴 🄸🄳🄴🄰

Does this count?

@Skillmon Of course, and there's no comma problem!

@CarLaTeX 20 points and I got a point-palindrome, too. That would be two on one day...

I did it!

two palindromes on a single day! (though I beg you all pardon for the miserable comma)

@Skillmon less than one day needed :)

@Skillmon yay

@FaheemMitha -- "The government used to be prohibited from specifying any specific manufacturer's product." When I was active on a standards working group, one of the other participants was representing the U.S. Government Printing Office and the IRS. He described in detail what was necessary to be able to obtain an IBM typewriter for the office -- every unique detail of the mechanism had to be specified, but the name "IBM" couldn't appear anywhere. Yes, this has apparently changed.

@barbarabeeton Same here in Brazil, you had to specify some components in order to narrow down to the specific brand, but not call by name...

@AlanMunn -- I have a grudge against Turbo Tax. When it first came out, my husband brought home a copy, so I tried plugging in the numbers already applied for the manual version, which was already finished and checked. Turbo Tax claimed that we were due a refund $3000 more than what I had calculated. We had already been audited once, and I didn't want to chance it again; I wouldn't be able to say where the numbers had come from. I don't trust computers -- been using them too long.

@Skillmon 4 palindromes without commas ;-)

@marmot what's that?

@Skillmon Four palindromes, as I said.

@marmot obviously, but which statistic is that from?

@Skillmon tex.stackexchange.com/users/121799/marmot?tab=tags

@marmot on that page I have 4 palindromes as well.

@Skillmon So 6 in total.

@marmot yes, but that's somehow boring. The other two are more interesting, imho.

How to donate to help rebuild Notre Dame: notredamedeparis.fr/friends/donate don.fondation-patrimoine.org/SauvonsNotreDame/~mon-don and nationalshrine.org/support-notre-dame

@marmot Here you have 52 palindrome numbers, not only 4 or 6 ;-)

@JouleV Good point. ;-)

@PauloCereda They do not have to rebuild, I have already seen it. ;-) (Just kidding!!)

@marmot oh

@Student404Mus DON'T SHOUT:-) that is a FAQ, you can not use left-right over a linebreak, use \bigl( on one line and \bigr) on the next.

@DavidCarlisle I've found a way to break hypdvips

@JosephWright you did want broken links

@DavidCarlisle Actually, scratch that: it's OK

@DavidCarlisle I do have one issue: you'll get a mail shortly

@JosephWright ooh

@PauloCereda Ive added you ...

@JosephWright ooh

@JosephWright <3

@HenriMenke I have one question: looking at line 56 of pgfmanual-en-tutorial.tex

\def\costhirty{0.8660256}

Why one has to do that when one can always use {cos(30)} later in the document, which is much more accurate IMHO?

or {sqrt(3)/2}?

ooh Endgame spoilers

@PauloCereda Ssh

@JosephWright ooh secrets

@PauloCereda Did you enjoy the PDF/PostScript mail?

@JosephWright It's quite... complicated yet amusing to debug. :)

@PauloCereda github.com/latex3/latex3/commit/… was a real pain!

@JosephWright ouch

@PauloCereda I did it once 'building up', but couldn't get it to work, so tried 'stripping down' the model I had .. only to find the same issue: objects :(

@JosephWright it seems to boil down to internal PDF specs... :(

@PauloCereda Not exactly

@JosephWright really?

@PauloCereda It's about how you save objects using pdfmark, as they get set up/_objdef { name }. The issue is I think that { ... } is normally a definition

@JosephWright Ah!

@JosephWright it looks as if the pdfmark syntax can simply not handle the situation that there are two objects.

@UlrikeFischer I'm not sure it's that ... I think it's a limitation on what one can save, or at least what I know how to save. I have a feeling cvx and exec are somehow needed

@UlrikeFischer Could you give me an example for the 'non-argument' links?

@JosephWright well I have no idea what hypdvips or you did. But I can see that the only difference between a failing and non-failing example in the ps is the /_objdef, and if one use quadpoints=true (and so get only one object) it works too.

@UlrikeFischer Really?? Doesn't for me!

\documentclass{book}
\usepackage{color}
\usepackage{lipsum}
\usepackage{hyperref}
\usepackage{xparse}
\ExplSyntaxOn
\makeatletter
\NewDocumentCommand \PdfMarkLink { m m }
  {
    \pdfmark[{#2}]
      {
        pdfmark = /ANN,
        linktype=url,
        Border=\@pdfborder ,
        BorderStyle=\@pdfborderstyle ,
        Color=\@urlbordercolor ,
        Action={<</Subtype/URI/URI(#1) >>},
        Subtype=/Link,
       objdef = foo
     }
  }
\NewDocumentCommand \ExplLink { m m }
  {
    \mode_leave_vertical:

@UlrikeFischer ^^^

@JosephWright what doesn't work for you (and I only tried the example in your mail).

@UlrikeFischer Example in my mail if I remove quadpoints = false

@JosephWright basically this here (in the original stored in l@section:

\documentclass{article}
\usepackage{hyperref}

\begin{document}

\section{abc}

%\newpage
\makeatletter
\hyper@linkstart{link}{section.1}some text\hyper@linkend

\end{document}

@UlrikeFischer I'll have to look at the hyperref code: I don't see why this can't get grabbed ...

@JosephWright ok after retesting: it works with quadpoints=true, but only if I remove the first short link.

@JosephWright there is no (real) problem to grab it in the specific case, the question is if grabbing is always possible.

@UlrikeFischer I really don't see any other way in the 'classical' route

@JosephWright if I do something like objdef = foo\name, and then give \name different definitions before the links (and use quadpoints=true) then it works for two links too.

@JosephWright and what is the problem? Do you need the width or the height of the link text?

@UlrikeFischer Yes, you've got to provide a border for the link as part of the link ... I guess one could save the data and dump it at the end, taking inspiration from hypdvips to do it all in PostScript ...

@UlrikeFischer pdfTeX can know the link size to write the /Rect entry

@JosephWright so basically you need to know the start and end position? How does hypdvips do it? Does it grab the link text?

@UlrikeFischer Well hypdvips sticks to the \pdfmark approach, so everything is grabbed. But the \Rect entry gets re-written. So if one used a dummy \Rect and stored the end point co-ordinates, you could back-up and change things in PostScript (at least I think)

@JosephWright does dvips has something equivalent to pdfsavepos?

@UlrikeFischer It has exactly \pdfsavepos

At least in pdfTeX/LuaTeX

@JosephWright oh, I hadn't realized that it works in dvi-mode too. If one knows the boundary of the link rectangle, can one set it "manually" by giving coordinates in dvips?

@UlrikeFischer pstricks was getting the position before pdftex had psfsavepos, if you don't need to feed the coordinates back to tex you can simply write a special that saves the coordinates in some postscript values, then you can use them on the same run within the postscript

@UlrikeFischer Like @DavidCarlisle says, TeX doesn't need these co-ordinates, only PostScript does. So it could be done ... I think

@UlrikeFischer What I'm still not seeing is the need, really: hyperref manages to use \pdfmark, which does need the text up-front

@JosephWright I seem to remember never really getting a good solution to what to do if you have the coordinates of the start and end points of the link but they end up being three lines or three pages apart, but my postscript is so rusty these days, I can't really remember anything

@JosephWright we are probably thinking in different levels. hyperref does open a \hbox in \hyper@linkstart and close it in \hyper@linkend, so it "grabs" the linktext. But the linktext isn't an argument, and so you can easily do something like \hyper@linkstart\bfseries\underline{some text}\hyper@linkend or put \hyper@linkstart in some command and the end in some other.

@UlrikeFischer That can't work with hdvips.def, surely?

@UlrikeFischer Hmm, OK, I see how it's done: mark the start of the shape, deal with the end inside \pdfmark. I'll give it some more thought.

@JosephWright why not? \hyper@linkstart{link}{section.1}\bfseries \underline{some text}\hyper@linkend works fine for me with dvips.

@UlrikeFischer Hmm, OK, I'll look to re-work the code again on this basis. I have a reasonable amount of stuff to start from, so a split will not be so hard

@UlrikeFischer Did you think about the attribute/target split? I'm not seeing it as useful

@UlrikeFischer OK, so what I'll do is re-work the stuff I have for dvips to be more in the style of pdfmark, plus split the other code back up (that's much easier)

@JosephWright on the code level I'm not sure, but on the conceptual level I think the disctinction is useful: on the one side one has attributes like color and border which affect a whole link class, and on the other side the specific action of a link.

@UlrikeFischer OK, if you feel it serves a purpose

@UlrikeFischer I've looked at the \hyper@linkstart code: it still starts a box in the dvips route, so we (I) have to think a little about how to keep the driver code 'sane'. I guess collect in all cases using the same method then unbox ...

@UlrikeFischer Actually, we are OK: with breaklinks active, there is no box. And we can do link breaking ....

@JosephWright on the driver level I'm not sure, I guess it depends if the goto name and thread variant are useful.

@UlrikeFischer I'll keep it for the present: we can always deprecate later if required/sensible

@UlrikeFischer I think goto name can be done using /DEST and a user link, not sure about thread

@JosephWright yes, if your new method works and gives more or less sensible link area, breaklinks should be active by default.

@UlrikeFischer It does, it's just a question of the object reference busines

@UlrikeFischer Tests thus-far suggest it works well, at present I've not provided a 'turn this off' option, though we may want to

\documentclass{book}
\usepackage{color}
\usepackage{lipsum}
\usepackage{hyperref}
\usepackage{xparse}
\ExplSyntaxOn
\makeatletter
\NewDocumentCommand \PdfMarkLink { m m }
  {
    \pdfmark[{#2}]
      {
        pdfmark = /ANN,
        linktype=url,
        Border=\@pdfborder ,
        BorderStyle=\@pdfborderstyle ,
        Color=\@urlbordercolor ,
        Action={<</Subtype/URI/URI(#1) >>},
        Subtype=/Link,
        objdef = foo
     }
  }
\NewDocumentCommand \ExplLink { m m }
  {
    \mode_leave_vertical:

@UlrikeFischer ^^^ Current test file

@UlrikeFischer Suggestions for more edge cases welcome

@JosephWright looks good. And unlike with pdftex there seem to be no problem with fancyhdr: the header doesn't get linked too.

@JosephWright on other news: I'm automating my install tools ^^^

@UlrikeFischer Which editor?

@UlrikeFischer Good!

@UlrikeFischer Good!

@JosephWright winedt. (That's a popup menu).

@UlrikeFischer Thought as much: still looks much as I remember it when it was my main editor

@UlrikeFischer It's funny: most of the current expl3 code is probably written in an editor for 'beginners' :)

@JosephWright ;-). For writing code I could use any editor probably. But for switching between texsystems and pdf viewers and such stuff, something with macros it really quite nice (and I have also a short cut to switch to the latex-dev versions (it is now unter "TeX Systems).

@UlrikeFischer Yes, much the same here

@JosephWright there is no equivalent for \pdfliteral~page yet in l3drivers right?

@UlrikeFischer Not yet, no: it's very low level and I was thinking just of \@@_literal_page:n or \@@_page:n

@UlrikeFischer I am working on all of this stuff! It's non-trivial

@JosephWright I was just asking ;-) I'm going through this driver and try to decide what I can already replace (I hope that some time it will change from a "luatex driver" at least to a "luatex + pdftex driver").

@UlrikeFischer I was planning, at least initially, not to provide public interfaces to 'raw' stuff like this. Do you have a specific use case in mind? Does it abstract at all?

@JosephWright the hyperref driver uses it, eg \pdfliteral~page{/OC/OCPrint BDC}, accsupp package too, and tagpdf anyway (I have to insert tons of BDC/EMC operators). I don't really see how to avoid to use it.

@UlrikeFischer OK, so \driver_page_literal:n?

That's the work of a few minutes to add for the next release (around the time TL starts updating again: will have links sorted by then)

@JosephWright what about the page/direct keywords? How would they fit in?

@UlrikeFischer Well we don't want keywords, hence for example I've said \driver_page_literal:n = \pdfliteral page {#1}. For direct, it's not come up thus-far (\pdfliteral covers most things I've wanted to do)

@JosephWright sorry didn't see the page, (I read it as pdf), but shouldn't it be \driver_pdf_page_literal:n?

@UlrikeFischer Is it a PDF concept? Doesn't it apply in DVI mode too?

@JosephWright no idea ;-) will have to check tomorrow. But from the description I always thought it is rather pdf orientated.

@UlrikeFischer Been looking at \hyperr@linkstart. In the dvips version, the limitation is that the link area height has to be assumed from \baselineskip. So one probably needs both versions (which is what hyperref does).

@UlrikeFischer I'll work on it tomorrow