Consider deprecating this driver

[stof]

This driver suffers from many cases where it does not behave properly (see failures in the testsuite).

It is also much slower than Selenium2 (see the time spent to run the testsuite), has less features than Selenium2, and is a pain to maintain.
I suggest marking this driver as deprecated (and unmaintained), acknowledging the fact that we haven't managed to fix it since years (the testsuite has never been green since the introduction of the driver testsuite ensuring proper driver behavior). If anyone is using the driver and wants to bring it uptodate, we can still make them a driver maintainer and undeprecate the driver if its testsuite becomes green.

If we do this, we should do a final release of the driver (as we did for SeleniumDriver).

/cc @aik099

[stof]

@aik099 what do you think ?

[aik099]

Since I don't like Sahi I'm ok with deprecating it.

Now we've deprecated 2 of the drivers already. Aren't we moving in the direction of having couple of drivers and ability (which might not be used by anyone) to add more drivers later. Not sure if it's a bad or good thing to have small driver count.

[stof]

Well, I'm looking into deprecating drivers which don't pass the driver testsuite. And the fact that it makes maintenance easier is a nice bonus.

the drawback of having a small driver count could be to design an abstraction which is too tied to the webdriver way and cannot be achieved elsewhere (I'm thinking about the discussion about alert/popup support here). Having multiple drivers is good to challenge our abstraction.
But I think of a way to avoid this issue: reject any addition to the Mink driver API if it cannot be implemented in at least 2 core drivers. So a webdriver-only API would then be rejected (and people would continue to access the underlying webdriver library to use it).

[aik099]

But I think of a way to avoid this issue: reject any addition to the Mink driver API if it cannot be implemented in at least 2 core drivers. So a webdriver-only API would then be rejected (and people would continue to access the underlying webdriver library to use it).

Yes, I was thinking about same thing, but wasn't sure how to approach it. For example minkphp/MinkSelenium2Driver#228 issue.

[aik099]

@stof , I guess we can deprecate driver when you're ready.

[stof]

I marked behat/sahi-client and behat/mink-sahi-driver as abandonned on Packagist and I updated the repo descriptions on github. So closing this