- From: PhistucK <phistuck@gmail.com>
- Date: Sat, 13 Sep 2014 23:07:34 +0300
- To: Doug Schepers <schepers@w3.org>
- Cc: Renoir Boulanger <renoir@w3.org>, List WebPlatform public <public-webplatform@w3.org>
- Message-ID: <CABc02_KaXnND=j6KM-Ls2YmkmFTBhWXacnX0=6+EWF9Nf=dFGg@mail.gmail.com>
Are we going to accept user contributions like "Chrome 37 - supported" (meaning, without tests)? And my other question still stands (though not urgent or actively relevant at this point, I still want to get the idea)... what if tests say a feature is supported in Chrome 37 and the user says otherwise (or the opposite)? How will we handle conflicts? I think I outlined my proposal at a high level. Are you looking for something even more concrete? If so, I asked before - where is the code that deals with writing and reading that file? After seeing the code or understanding the process and the design, I can try and come up with a concrete plan and execute it if it is approved. ☆*PhistucK* On Sat, Sep 13, 2014 at 9:37 PM, Doug Schepers <schepers@w3.org> wrote: > Hi, PhistucK– > > In addition to what Renoir said… > > On 9/12/14 8:27 AM, PhistucK wrote: > >> Looks reasonable. >> Thank you! >> >> Perhaps off topic, but regarding the data itself - >> 1. Is there a scheduled task that updates the data from the various >> sources (only MDN at the moment, right?)? >> > > No, there is no scheduled task yet. > > We don't plan to retrieve data from MDN again, for a few reasons: > > 1) The MDN data is not well-structured for extraction; I think they are > working on this, but for now, it's not trivial to get the data, and it > requires a good bit of post-processing; > > 2) Once we have the MDN data, and have normalized it, we still have to map > the naming and category conventions to those of our own pages and site > structure, which is not 1:1; > > 3) There is no regular update of their data, such that we could retrieve > it on a schedule; this might change in the future, since they are working > on their compatibility information; > > 4) The MDN data is not necessarily accurate; the intent is certainly > there, but their compatibility results are not based on tests, but only on > the judgment of their contributors; and where those contributors may used > tests, the tests are not exposed to the reader for verification of the > quality of the test; > > 5) The MDN data is not detailed enough; though they have some subfeatures > listed, it's not systematic or consistent, and doesn't typically cover edge > cases, or combination with other features. > > The MDN data was an excellent basis for a first start at our compat info, > and we're glad it was there; we're grateful that they allowed us to use it. > > CanIUse.com has also offered the use of their data, which is more > test-driven (and thus more accurate), is easily extracted, and which does > have regular updates; however, at a feature level, it is organized more > broadly than our pages, so the mapping between our page/category hierarchy > and CanIUse is even more challenging then that of MDN. Still, it might be > worth the effort, because of the other advantages. > > Ideally, we will use the W3C test-suite results data [1]. That also needs > some normalization and adaption, but the value is high because it's > extremely detailed and accurate information. > > > a. If so, are conflicts handled (user added data using pull requests)? >> How? >> b. If not, how is the data kept up to date and synchronized? >> > > These will be issues, but not for MDN data, as mentioned above. > > > 2. The data-human.json file is huge and cannot be edited in the web >> interface of GitHub as a result. Can you split it to folders by topic >> and files by names (or any other way that creates small files that can >> be easily edited)? During the build (or whatever it is that processes >> the data), everything could be combined and then processed. >> > > That seems like a good idea. Do you have a proposal to handle the split > (and the merge)? > > > [1] https://github.com/w3c/web-platform-tests > > Regards- > -Doug > > ☆*PhistucK* >> >> On Fri, Sep 12, 2014 at 4:31 AM, Renoir Boulanger <renoir@w3.org >> <mailto:renoir@w3.org>> wrote: >> >> Hi all, >> >> I am about to push an update[1] on our wiki and I thought i’d ask for >> text validation. >> >> There’s nothing to see on the wiki or staging —I had no time to work >> on >> a staging server yet— but you can look at the screenshots in the >> issue [0]. >> >> >> >> Wanna help with the text?: >> >> * When no data found: >> >> [[ >> There is no data available for topic "%s", feature "%s". If you think >> that there should be data available, consider <a href="%s">opening an >> issue</a>. >> [[ >> >> >> * When data found, to give advice how to help: >> >> [[ >> Do you think this data can be improved? You can ask to add by <a >> href="%s">opening an issue</a> or <a href="%s">make a pull >> request</a>. >> ]] >> >> >> >> [0]: https://github.com/webplatform/mediawiki/issues/17 >> [1]: >> https://github.com/webplatform/mediawiki/compare/compatables-update >> >> -- >> Regards, >> >> Renoir Boulanger | Developer operations engineer >> W3C | Web Platform Project >> >> http://w3.org/people/#renoirb ✪ https://renoirboulanger.com/ ✪ >> @renoirb >> ~ >> >> >> > >
Received on Saturday, 13 September 2014 20:08:41 UTC