- Default output is PDF (DVI is a port on my computer :) )
- Optional DOC/DOCX output (even if it supports just a subset of the program's capabilities)
- Bibliography management built-in (no separate ancient BibTeX)
- Command-line searching and importing from Pubmed / Google Scholar / etc. built-in
- Bibliography styles can be created/edited by normal humans
- Syntax closer to Markdown or reStructuredText, to enhance readability of source file (and maybe shrink the learning curve)
- Program is smaller than 1.6 GB (the size of the default MacTeX distro).
I've played with writing something like this for a while (in Python). The hardest part for me has been dealing with bibliography data - there's always one more weird Pubmed record that breaks my XML parser. Makes me really appreciate Knuth/Lamport/Patashnik and their successors.
- The base .tex syntax will probably never change, but there is no reason why some tool/plugin/luatex extension couldn't read something else and convert as necessary. I used Deplate and Pandoc for a while.
- AFAIK ConTeXt has some integrated replacement for bibtex
I don't think we need a successor to LaTeX. The base is solid and ready to be built on, especially now with a sane programming language in the core of LuaTeX.
Pdftex (the default Tex engine) supports this. Both Xetex and Luatex do as well, and both of these additionally assume UTF8 as the input encoding and allow you to use True Type and OTF fonts directly.
2. Optional DOC/DOCX output (even if it supports just a subset of the program's capabilities)
igneous4 mentioned Pandoc, which supports RTF as an output format, which is as good as doc/docx for importing text with basic formatting into Word. Pandoc also supports Context. If you want to move the other way, there is docx2tex, which is the best Word to Latex converter.
3. Bibliography management built-in (no separate ancient BibTeX)
I think there's nothing satisfactory here for Latex. You might like to look at Jabref. With Context, the automatic build process makes the invocation of Bibtex invisible. There are many make-like tools for Latex, none are seamless.
4. Command-line searching and importing from Pubmed / Google Scholar / etc. built-in
Again, nothing built-in, but Zotero will do what you want.
5. Bibliography styles can be created/edited by normal humans
Again, nothing adequate here. Try looking at Biblatex. I hand/regex edit .bbl files if I have special bib style needs.
6. Syntax closer to Markdown or reStructuredText, to enhance readability of source file (and maybe shrink the learning curve)
Tex's markup is actually nice. I often wish I could use Texisms when I am editing Markdown, etc. Pandoc allows you to convert between all of these.
7. Program is smaller than 1.6 GB (the size of the default MacTeX distro).
Most of which is fonts. If you have less than voracious font needs, or use TTF/OTF (via. Luatex/Xetex), you can get away with a small fraction of this. Try a more lightweight Texlive installation, and find out what you need.
The hardest part for me has been dealing with bibliography data - No surprise. Bibliography management used to be a relative strength of the Tex family, because the competition was so extremely shoddy. Now it is a weakness; Word 2010 with Endnote is actually very nice here. It's a shame that relatively few Latex hackers appreciate how far behind Tex&co have fallen.
Makes me really appreciate Knuth/Lamport/Patashnik - Knuth is inspirational, his code keeps on rewarding those who read it. Patashnik doesn't get much love from me: he held so many wrong opinions vehemently.
Adding images to Latex documents is straightforward using the core libraries graphicx and putting \includegraphics{FILENAME} at the place you want the graphic. Adjusting placement to left, right, overhand, etc., is done using Tex primitives. It's true that using Tex , Latex and the main Latex packages correctly is not straightforward and that until you have mastered them, you can get into difficulties, but that is also true of Word.
The difference is that it takes longer learn how to produce any sort of ugly document in Latex than in Word, and longer to produce beautiful documents in Word than in Latex.
Sorry, but I'll stick to my point. We should stop making excuses for software that was great in its time but is simply not adequate anymore by today's standards.
Almost nothing is straightforward in LaTex, and images aren't either. Placing an image exactly where you want it can become a serious issue - even more so in documents that change and get reflowed all the time.
Yes, LaTex often makes good decisions that require little tweaking. But the moment you disagree with one of its decisions you find yourself in a world of pain.
Almost nothing is straightforward in Word, and images aren't either.
Word works very intuitively until you try to achieve fine control of layout. Then if you haven't done things the correct way, where Word's document model is not more intuitive than Latex's, getting the results you want will be painful.
It is true that the distance between Word and Latex in terms of suitability for producing high-quality output has narrowed drastically in the time from, say Word 6.0 to Word 2010. But the Tex family is not software that was great in its time. I use both Word and Tex-based technologies most working days, and I massively prefer to the Tex-based technologies in terms of quality output. It is still the case that for obtaining quality proofs from a .doc/.docx file, professionals use external software such as Adobe Indesign. Indeed, there are workflows that take Word documents and convert them to Tex-based representations for typesetting.
I don't know a piece of free software that directly competes with LaTeX, the latter has been the de-facto standard for university papers and a small group of book-writers/enthusiasts for about 20 years.
However, pitting LaTeX against Word is about as meaningful as pitting Linux against Windows. You wouldn't excuse the thorny Linux desktop experience with the security track record of windows either - it'd be an apples vs oranges comparison.
Open Office exists as part of a free Java-based ecosystem. I couldn't say it is used in any serious publication workflows, but the components are there. Context is aimed at serious publishers, and has strong roots in educational (i.e., not academic) publishing.
Apples vs. oranges: the point of this subthread has been that I thought your statement But the underlying toolchain and language are past their expiration date. By far. and backing this up by talking about how hard "easy" things can be that turn out to not to be trivialities anywhere. Latex is not for casual users, but it is very much not past its expiration date for authors who are serious about typesetting. I'm not making excuses for Latex, nor am I unfairly bashing the alternatives.
:-) TBH, though, there's just no comparison for producing nicely-formatted research papers. Especially in that eye-straining, dual-column, 9pt ACM conference format.
Of course, there are still some ugly corners, and the underlying code isn't always pretty. Things like Luatex sound quite promising in this regard.
Pdftex is currently the standard Tex engine for the two main Tex distributions, Texlive and Miktex.
Yes, the output looks \bf{gorgeous}. But the underlying toolchain and language are past their expiration date. By far.
Tasks as simple as "let's add an image here" or "let's move this to the left a little" frequently turn into hour-long journeys deep down into the guts of a 30 year old codebase.
There is no reason why a modern tool couldn't use the same algorithms to create equally beautiful output. We're not writing raw postscript to talk to our printers anymore either.
\end{rant}