Move Mosquito Server Sent Events tests

[odinho]

No description provided.

[dontcallmedom]

I've started to look at these — they're not quite ready for merge.

Among other things, the tests currently rely on a server-side component run by node.js to send events as needed; is that acceptable? If not, what is the proper way to define the server-side aspects of these tests?

[Ms2ger]

Eventually, the "proper way" is intended to be a simple python server, which does not yet exist. For now, anything that runs on w3c-test.org is fine, I guess.

[odinho]

It will probably be better to look at Opera's server sent events tests #60, because they will work on w3c-test.org (using PHP). They will need some quick checkups that they are sane wrt. the spec. We had a contributor do some fixes there already, and I have reviewed and accepted those changes. But the other files are left.

[dontcallmedom]

Having compared the MOSQUITO tests with the opera tests, the Opera tests cover all the things the MOSQUITO tests do, and are much closer to be ready than the MOSQUITO ones, so I would suggest to drop the MOSQUITO tests.

[cynthia]

On top of that, the "server component" is not really functional. It has interesting bugs like 1) append .html on incoming requests for resources, so if you ask for "/user-agent-announce-connection.html" it'll try to send you "user-agent-announce-connection.html.html", which 2) crashes because it doesn't handle missing files and 3) will crash even if you ask for "/user-agent-announce-connection" because the tests look for files in resources/, which are at least not in this subtree. (Unless I'm missing something, but I would expect a testsuite to be a bit more plug-and-play.) So at least the server component needs to be fixed or re-written. (and 4) it will send HTTP 200 even if the file doesn't exist, although it crashes immediately after that so I guess that's fine. :P)

[odinho]

Sounds good to me that, -- although I guess we shouldn't just close it. Rather fix and integrate the Opera tests, and then when we actually already have that submission in this one won't be needed.

And maybe we should also ping the mailing list? I guess not everyone is following these pull requests. I don't know if we have dropped anything like that before? I did refuse a pull request to the Opera SSE tests because it was wrong, and redid the correct parts of it after-the-fact, but that was a bit different :-)

[AFBarstow]

@velmont @dontcallmedom - I believe Tina's analysis is consistent with what I think you two are saying -> WebApps should consider Opera's SSE tests as sufficient to test the SSE Candidate and the MOSQUITO tests should not be used. That is probably why she only included Opera's test in the SSE test results doc she created http://www.w3.org/wiki/Webapps/Interop/ServerSentEvents.

Odin - perhaps you should confirm that on public-webapps-testsuite? Also, how do RfRs map into this brave new GH world? I think the requirement is a clear request to the WG to review tests we want to use for CR interop testing.

[odinho]

Removed Mosquito tests in favour of newly moved Opera tests.