OK, here are my belated replies throughout this thread, combined into
one message! I'll start with my overall conclusions and then give
replies to individual messages, all concatenated below.
I would be quite happy to reconceptualize the Ur project site as a
community project. For me to it would be ideal for a set of non-me
volunteers to handle both the design and the hosting. It would probably
be best for me to register a domain for the site (the impredicative.com
part hasn't seemed like a good fit for a while) but point hostnames as
directed by volunteers.
So, the big question: who would be interested in taking charge of some
substantial part of that kind of community effort?
All of that might make sense to do simultaneously with a switch to
GitHub, or maybe the GitHub switch deserves to come earlier. One task
that makes sense to tackle concurrently with a GitHub switch is a
redesign of the Ur/Web extended standard library, which currently is
split across several Mercurial repositories. Any thoughts on whether it
would be better to have one Git repository for all of the extended
standard-library content listed as "Officially Blessed" here?
http://www.impredicative.com/ur/libraries.html
Would it make sense to maintain separate Git repos for all of them, but
maybe also provide one repo that includes all the others as submodules?
(I'm also thinking of removing the gui and meta libraries in favor of
migration of appropriate code into UPO.)
On 07/28/2015 04:54 PM, Torstein Saltvedt wrote:
On Tue, Jul 28, 2015 at 7:32 PM, Adam Chlipala <[email protected]
<mailto:[email protected]>> wrote:
Thanks to everyone for their thoughts on the project web site.
Here's my summary of the 5-ish opinions expressed:
- It's not clear that the large background graphic from Torstein's
design is the way to go. Somehow it may be out of keeping with
the "character" of the Ur/Web project.
Aesthetics are subjective, and although I think the current site looks
outdated, it's only one of the reasons for the redesign and arguably
the least important one.
Those are definitely some compelling reasons (not quoted here)!
Unfortunately, they draw largely on web-design knowledge that I haven't
bothered to develop myself. But the plan above can help get around that
problem.
- Moving to Git & GitHub also seems like a no-brainer. One
concern about GitHub was expressed, regarding censorship. I
personally am not too worried there, as it's easy to maintain
"mirrors" of a Git repository all over the place, to be ready in
case one main provider goes over to the dark side. The pros seem
to outweigh the cons, considering how many potential contributors
already have GitHub accounts and are used to using GitHub.
Does this mean that the project will adapt the GitHub Wiki, Issue
tracker and the Pull Request model as well? Will you still accept
patches through this mailing list?
Yes, I'm suggesting normalizing the Ur/Web community process to be the
standard GitHub one, which I think then replaces all of the current
infrastructure beyond the mailing list and the general web presence
(including demos & documentation).
On 07/27/2015 08:30 PM, Stefan Scott Alexander wrote:
Also, "eating your own dog food" would probably be a plus. It
only makes sense that a website for a web programming language
should be programmed in the language itself.
I'm not sure about this one. Ur/Web is for web _apps_, not web
_sites_, so it may be a mismatch for a largely static site.
Agreed, but it would still make for a great demo.
Right, so, for that reason, I could see being convinced that it's a good
idea.
On 07/28/2015 05:59 PM, Sergey Mironov wrote:
2015-07-28 20:18 GMT+03:00 Adam Chlipala <[email protected]>:
Sergey, anyone else out there: might I be able to interest you in
implementing and/or hosting such a service? I would be very happy to link
to it prominently from the Ur project site!
I am quite busy at the moment. Probably, I'll have some time next
month to address the problem if nobody wants to do it.
For inspiration, one may checkout my old half-done project
urweb-pastebin.
Neat, thanks. One other person indicated interest in building a
prototype system some time soon, and your code may indeed be helpful as
a starting point.
On 07/28/2015 10:19 PM, Timothy Beyer wrote:
The site needs to look trendy, so things like bootstrap widgets would
be important in promoting Ur/Web. Plus, a menu that toggles between
desktop or mobile menu (ex. SmartMenus [2] with Bootstrap 3 addon [3])
also looks really modern, although it might be a bit overkill for this
project.
I think I'd rather replace "trendy" with "good" in your description
above. ;) I wouldn't mind a Bootstrap-based design, though. I'm
already using the Ur/Web Bootstrap library in production apps via UPO.
Possibly a prerequisite for the above, but it would be cool if we had a webhost
available with Ur/Web and PostgreSQL installed so that it would be easier to
host Javascript style interactive demos. These types of demos could really
make a big selling point for Ur/Web, especially to newcomers and those
skeptical of its practical uses.
Well, it's trivial to do with a Linux virtual machine, and it will be
even more trivial once Debian and Ubuntu do stable releases that include
the new Ur/Web packages.
I agree that it couldn't hurt to make it all easier.
Not directly related to github, but we should implement support for
syntax highlighting libraries such as highlight.js (perfect for
webpages and markdown), pygments (for TeX/LaTeX).
I agree that it's a good thing to add Ur/Web support in as many
supporting tools as possible! I don't see working on it myself (I
focused my efforts on the Emacs mode's syntax highlighting), but I'd be
glad to accept patches into the main Ur/Web source tree.
If we are going with github, we should also have a "Fork me on github"
ribbon on the top right of the page.
Yeah, sounds like a good idea.
Another very useful thing would be a tutorial that doesn't assume ML and
Haskell familiarity, ideally written by someone beside me, since
Ur/Web's design has been in some sense optimized for my brain. :) Any
takers there?
I would like to make some tutorials at some point, although I'm curious which
tutorials people would like to see?
Yeah, that's a good question, and it's one that I'm not so qualified to
answer, re: my general contention that Ur/Web documentation will get
stronger as it's increasingly written by people who didn't design the
language, whose likely confusion points I'm especially unqualified to
predict.
Right now, I'm focusing on library stuff, rather than tutorials, such as GUI
components similar to what I've used on big Javascript projects, and JSON RPC.
Can you elaborate on the JSON RPC part? I can't think of anything of
that kind that wasn't already well-supported by Ur/Web and its extended
standard library 5 years ago.
On 07/30/2015 05:24 PM, Istvan Chung wrote:
I'd like to express a somewhat late -1 vote for moving to git. Git has
a number of technical deficiencies compared to Mercurial. [...] All in
all, the main reason a project wouldn't want to use Mercurial is
because it has some barrier to entry due to being less well-known than
git. [...] The main advantages associated with GitHub are a nicer web
interface, issue tracking, and pull requests.
I agree that Mercurial is technically superior to Git, and I believe
that holds true for both novice and expert users. The reasons for
switching to GitHub are mostly social: GitHub is now the standard
platform for open source, and it's likely that potential Ur/Web
contributors would be unfamiliar with Mercurial and do "cost accounting"
that assigns responsibility for learning Mercurial to the Ur/Web
project, which reduces their motivation to start contributing. It's
also true that pretty much everyone has a GitHub account by now, and it
reduces the barrier to entry, for contributing code or reporting bugs,
to be able to use a standard GitHub account. Barriers to entry are also
reduced by allowing the standard GitHub tools to be used, in place of
others like Mantis that will also require from-scratch learning for most
folks. (I also think that GitHub's issue tracking is probably
substantially superior to Mantis, just because of how much simpler the
former interface is.)
Overall, as for the web site design, it's misleading to analyze the
options by polling the current contributors, since there might be a much
larger population of potential contributors who were turned away by the
present configuration!
_______________________________________________
Ur mailing list
[email protected]
http://www.impredicative.com/cgi-bin/mailman/listinfo/ur
