Timeline for EOL of Firefox 3.x support

[Simon Stewart]

Hi,

Mozilla no longer support Firefox 3.x, and we're now on the second ESR of the new era of rapid updates. Looking around, browser market share for Firefox 3.x has slipped below 2% (at least according to StatCounter). It's time we had a serious discussion about dropping webdriver support for 3.x.

So, questions:

1) Should we drop support for 3.x?
2) if so, when and how?
3) if not, why? And when do we stop supporting it?

I have opinions and ideas, but don't want to skew the conversation just yet.

Cheers,

Simon

[Adam Goucher]

To go meta a second, Se is more than just WebDriver.

Should the discussion be whether 'Se' drops support for 3.6.x 'as a
project'? Se-IDE still has 3.6.x as its minVersion. And there is a
Firefox2Locator.java, Firefox2Launcer.java, Firefox3Locator.java,
Firefox3Launcher.java, which are Se-RC so there is at least a bit of FF
version-y stuff in Se-RC as well.

But to answer the specific questions,

1) Yes
2) Announce on se-dev, se-users, webdriver and the blog, wait a month,
release with removed support, announce again

-adam

[Simon Stewart]

Yes, you're right. This should be for the whole project. I left anything based on Core out of the equation because it's in maintenance mode --- I have no desire to make radical changes to it (such as deleting all those Firefox2 and 3 launchers) --- but we can include it in this discussion.

Simon

--

[David Burns]

1) yes
2) now
3) if people suggest no they can easily fork and maintain it themselves. Mozilla has been doing rapid releases for nearly two years and we have sent messages to those still on it.

Note: when WebDriver API switches to Marionette you get what you are given just like ChromeDriver and OperaDriver so there isn't much choice...

David

> --
>
>

[David Burns]

Just remembered from the last discussion on this that I started a couple months ago we were waiting on Google updating their infrastructure. 

David

--
 
 

[Simon Stewart]

Indeed. That's one of the reasons for phrasing things the way I did :) Jason and Daniel?

Simon

--
 
 

[Daniel Wagner-Hall]

On 5 January 2013 15:04, Simon Stewart <simon.m...@gmail.com> wrote:

1) Should we drop support for 3.x?

Definitely.

 

2) if so, when and how?

We [Google] should be fully free of any firefox <10 at the end of the quarter, but as long as we [Selenium] don't go out of our way to break things between now and then, we [Google] shouldn't have too much hassle.

How about we:

* Announce a formal deprecation somewhere

* When we drop support for ff10 bump the minVersion to 17 [might as well combine the major version deprecations :)]

* After that, be happy to add new code which depends on ff10+, but not go out of our way to remove old stuff unless it's actively in the way of development or causing bugs.  With Marionette on the horizon, doing any major pieces of surgery (e.g. trying to do a sweep for now dead code) to the FirefoxDriver seems riskier than it is worth.

FYI Jason's on holiday for the next couple of weeks; I'm not sure how much online he's going to be, but he may not see/respond to this quickly.

[Simon Stewart]

Inline.

On Sun, Jan 6, 2013 at 1:14 AM, Daniel Wagner-Hall <dawa...@gmail.com> wrote:

On 5 January 2013 15:04, Simon Stewart <simon.m...@gmail.com> wrote:

1) Should we drop support for 3.x?

Definitely.

 

2) if so, when and how?

We [Google] should be fully free of any firefox <10 at the end of the quarter, but as long as we [Selenium] don't go out of our way to break things between now and then, we [Google] shouldn't have too much hassle.

Given what I know of the Google infrastructure, could you snapshot the xpi?

 

How about we:

* Announce a formal deprecation somewhere

If we decide to do this (and it looks like we do), I'll write a blog post.

 

* When we drop support for ff10 bump the minVersion to 17 [might as well combine the major version deprecations :)]

* After that, be happy to add new code which depends on ff10+, but not go out of our way to remove old stuff unless it's actively in the way of development or causing bugs.  With Marionette on the horizon, doing any major pieces of surgery (e.g. trying to do a sweep for now dead code) to the FirefoxDriver seems riskier than it is worth.

These two statements don't mesh. It sounds like what you really want to happen is for us to bump minVersion to version 10. I agree about avoiding a wholesale rewrite of the firefox driver.

 

FYI Jason's on holiday for the next couple of weeks; I'm not sure how much online he's going to be, but he may not see/respond to this quickly.

Google's a data driven company. I'd suggest identifying the heaviest current users of Firefox 3.x within the company and seeing how much traffic on their live sites are from 3.x. But you know this already :)

I think we're close to making up our minds on this,

Simon

[Kevin Menard]

In a separate thread we discussed supporting ESR and ESR-1.  This would mean setting minVersion to 10, not 17.

--
Kevin

Daniel Wagner-Hall

Saturday, January 05, 2013 8:14 PM

--
 
 

Simon Stewart

Saturday, January 05, 2013 10:04 AM

Hi,

Mozilla no longer support Firefox 3.x, and we're now on the second ESR of the new era of rapid updates. Looking around, browser market share for Firefox 3.x has slipped below 2% (at least according to StatCounter). It's time we had a serious discussion about dropping webdriver support for 3.x.

So, questions:

1) Should we drop support for 3.x?

2) if so, when and how?

[Simon Stewart]

OK, so my conclusion from this thread:

1) I will announce the plan to drop Firefox 3.x support on selenium-users, webdriver lists and the blog.

2) Given that there is no strong pushback that can't be addressed, drop official support for 3.x in selenium 2.32

3) Restate our official position for Firefox as ESR, ESR-1, N, N-1

Simon

--
 
 

[David Burns]

One thing we should note that all of this is pre-Marionette.

ESR N -1 (currently Firefox 10) is about to EOL in the next month or so and has no Marionette code in it and neither will Firefox 17 ESR since the non-debug version of Marionette isnt available to the masses.

A Mechanism has been agreed in principle for us to move to opt-builds and we just need the code for it. In return we are giving more Mobile support in the future. Now before people start running to the hills saying that Mozilla is being unreasonable this is how other browser vendors do things for their browsers that the project doesnt look after

In the Marionette world, support is in every version straight away unlike now where we play catch up. Marionette is part of the Firefox test infrastructure and you can see tests being run (https://tbpl.mozilla.org/?tree=Mozilla-Inbound&jobname=marionette)

David

David Burns
URL: http://www.theautomatedtester.co.uk/

--