Re: ReSpec 3.2

FYI, I just converted the WebVTT spec successfully to the new version
- and found some bugs along the way. ;-)

I also moved my bug-assist box a bit down, so it would not conflict
with the respec box. Incidentally, it would be a nice idea to
integrate bug-assist into respec.

Nice work!

Cheers,
Silvia.


On Fri, Oct 18, 2013 at 11:19 PM, Robin Berjon <robin@w3.org> wrote:
> Hi all,
>
> if you're a ReSpec user, please pay attention to this message!
>
> I am about to release ReSpec 3.2, which is a fairly major change. I've run a
> fair number of tests (from the test suite and additional ones) and it
> appears to be behaving well, but you never know.
>
> I initially wanted to make an RC at a different URL, but the extent of the
> changes are such (notably the new UI stuff and how it is lazily loaded) that
> that's not easily possible.
>
> After the release I have a list of specs that I know to be using ReSpec that
> I will go through to triple check that all is fine. But just in case, I'd
> appreciate if you could check at your end too. At the first sign of trouble
> I'll back the changes out.
>
> Such major changes are rare (the last one was 3.0, 20 months ago). The next
> ones in the 3.2 series will be small and incremental as usual.
>
> Things that have changed:
>
>   • The most user-visible change is that there is now a UI in the top right
> corner. It features a button that when pressed shows a drop-down menu of
> things that can be done. Right now that list is limited but the modules that
> implement the functionality in the drop-down are loaded lazily which means
> that we can add functionality there without worrying about code size
> (notably linting).
>
> Next to the ReSpec button are two pills that appear when there are errors
> (red) or warnings (orange). Clicking them opens the list of issues that
> ReSpec has noticed while running.
>
> Error detection has been improved a lot in conjunction with this. This ought
> to help with the many cases of "it isn't working" bug reports when the error
> messages used to be in the JS console.
>
> In addition to saving, which is still accessible as Ctrl-Alt-Shift-S, errors
> can be opened as Ctrl-Alt-Shift-E and warnings as Ctrl-Alt-Shift-W.
>
> Overall the UI can be made prettier — input welcome. I also suspect that it
> may have some A11Y issues which I want to iron out ASAP.
>
>   • If the browser supports it, the saving dialog now uses the download
> attribute on <a> to trigger a real download of the generated output, as
> "Overview.html". For browsers that don't support that yet, you can
> nevertheless right click the button and save link as. In both cases this is
> faster than the previous methods.
>
>   • Code size is down another 10K, now at 44% of the original.
>
>   • The legacy module has been completely removed (work that started with
> 3.0). All of the useful functionality in it is now available in proper new
> generation modules, most notably a new core/biblio module.
>
>   • We no longer need to do weird things with JSON-P and <script> elements,
> all is loaded cleanly as JSON. In theory this means that we should be able
> to make the biblio parts fully asynchronous (which could have a very
> positive impact on perceived performance).
>
>   • The functions that generate snapshots are now exposed on core/utils and
> can be called at will.
>
>   • simple-node, a small library I wrote from the days before jQuery, has
> been almost eliminated. It is now only used by core/webidl-oldschool. I also
> removed everything from it that wasn't needed for that (e.g. namespace
> support).
>
>   • Note that several of the above changes combined completely eliminate the
> "berjon" (JS) namespace that used to exist. If you have some really dirty
> hacks that tap straight into the internals that's the first error you're
> likely to get.
>
>   • We've upgraded to the latest RequireJS (notably for building). More
> notable, we now use jQuery 2.0. That drops support for some of the older
> IEs; I don't think it'll be an issue for this usage.
>
>   • Testing has been made more correct in several cases.
>
> Share and enjoy!
>
> --
> Robin Berjon - http://berjon.com/ - @robinberjon
>

Received on Monday, 28 October 2013 00:36:52 UTC