Fennec User Agent

Mark Finkle

unread,

Jan 31, 2012, 9:11:17 AM1/31/12

to

After much deliberation, a decision has been reached on the Fennec UA
string. Fennec will use the following UA for phones:
"Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131
Firefox/12.0a1 Mobile/12.0a1"

Differences from current UA:
* Changes "Fennec" to "Mobile"

And the following UA on tablets:
"Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131
Firefox/12.0a1"

Differences from current UA:
* Drops the "Mobile" (nee "Fennec") token

As you can see, this is a very minimal change from the current UA, and
does not have some of the changes discussed over the last weeks,
summarized here:
https://wiki.mozilla.org/Fennec/User_Agent

After looking at the data collection results
(https://wiki.mozilla.org/Fennec/User_Agent#John_Jensen.27s_Data) and
some screenshot data, we determined that we could make significant
improvements to downloaded content quality without making large
mobile-only changes to the UA.

Other changes listed in the Wiki doc regarding finger-printing and cruft
cleanup are common to Mozilla in general, not just Fennec. As Mozilla
changes those tokens, mobile will "inherit" the changes as well as desktop.

See this bug for the patch to make the changes:
https://bugzilla.mozilla.org/show_bug.cgi?id=671634

Henri Sivonen

unread,

Jan 31, 2012, 9:46:56 AM1/31/12

to dev-pl...@lists.mozilla.org

On Tue, Jan 31, 2012 at 4:11 PM, Mark Finkle <mark....@gmail.com> wrote:
> After much deliberation, a decision has been reached on the Fennec UA
> string.

For future reference, it would be useful to know how these decisions
get made so that people don't expend energy uselessly on the wrong
forums.

> Fennec will use the following UA for phones:
> "Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131
> Firefox/12.0a1 Mobile/12.0a1"

Why is there a "/12.0a1" part after "Mobile"?

> And the following UA on tablets:
> "Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131
> Firefox/12.0a1"

Considering that some Web authors seem to treat tablets and desktops
as substantially different device categories, won't this lead to
sniffing for "Android" in order to target tablets making it harder
(compared to the proposal if having a clear "Tablet" token) for
Mozilla to get tablet content if/when Mozilla launches Firefox on
non-Android tablets? (E.g. hypothetical B2G tablets in the future.)

> Other changes listed in the Wiki doc regarding finger-printing and cruft
> cleanup are common to Mozilla in general, not just Fennec. As Mozilla
> changes those tokens, mobile will "inherit" the changes as well as desktop.

You could drop the "Linux armv7l" part on mobile even if the analogous
part is staying on desktop. Especially if the "Mobile"/"Tablet" token
was placed in that slot as seemed to be the consensus earlier.

(I think it makes sense to keep the Gecko/??? token in sync between
desktop and mobile.)

--
Henri Sivonen
hsiv...@iki.fi
http://hsivonen.iki.fi/

Dao

unread,

Jan 31, 2012, 9:52:26 AM1/31/12

to mark....@gmail.com

I agree that keeping desktop and mobile in sync is good in principle,
although, I as I understood it, the gathered data suggested that it
didn't really make a difference for compat, so mobile could easily lead
and desktop could follow.

I see no reasoning whatsoever for adding "Mobile" as a product token and
with the Gecko version number duplicated. Can this be changed back to
what https://wiki.mozilla.org/Fennec/User_Agent suggests based on John's
data?

dao

Ben Bucksch

unread,

Jan 31, 2012, 10:02:53 AM1/31/12

to

On 31.01.2012 15:11, Mark Finkle wrote:
> After much deliberation, a decision has been reached on the Fennec UA
> string. Fennec will use the following UA for phones:
> "Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131
> Firefox/12.0a1 Mobile/12.0a1"

That's strange. I thought we had the discussion here, and the outcome
was something else, as Gerv posted. How was this "deliberation" and
"decision" reached?
A consent on this newsgroup *is* the project decision. Gerv also is the
owner of the UA string spec.

Mark Finkle

unread,

Jan 31, 2012, 11:13:29 AM1/31/12

to Henri Sivonen

On 01/31/2012 09:46 AM, Henri Sivonen wrote:
> On Tue, Jan 31, 2012 at 4:11 PM, Mark Finkle<mark....@gmail.com> wrote:
>> After much deliberation, a decision has been reached on the Fennec UA
>> string.
>
> For future reference, it would be useful to know how these decisions
> get made so that people don't expend energy uselessly on the wrong
> forums.
>
>> Fennec will use the following UA for phones:
>> "Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131
>> Firefox/12.0a1 Mobile/12.0a1"
>
> Why is there a "/12.0a1" part after "Mobile"?

The HTTP 1.1 spec requires a "/" in product tokens

>
>> And the following UA on tablets:
>> "Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131
>> Firefox/12.0a1"
>
> Considering that some Web authors seem to treat tablets and desktops
> as substantially different device categories, won't this lead to
> sniffing for "Android" in order to target tablets making it harder
> (compared to the proposal if having a clear "Tablet" token) for
> Mozilla to get tablet content if/when Mozilla launches Firefox on
> non-Android tablets? (E.g. hypothetical B2G tablets in the future.)

It's still a bit early for tablets and we don't really know how webdevs
are going to target tablets specifically. I'd rather wait to see how
this evolves before adding tablet-specific tokens.

>> Other changes listed in the Wiki doc regarding finger-printing and cruft
>> cleanup are common to Mozilla in general, not just Fennec. As Mozilla
>> changes those tokens, mobile will "inherit" the changes as well as desktop.
>
> You could drop the "Linux armv7l" part on mobile even if the analogous
> part is staying on desktop. Especially if the "Mobile"/"Tablet" token
> was placed in that slot as seemed to be the consensus earlier.

We could, and we might - but it's not pressing right now. We also need
to consider the impact of x86 Android based devices. Hopefully, the
impact won't be such that the UA needs to carry this data.

Dao

unread,

Jan 31, 2012, 11:23:34 AM1/31/12

to Mark Finkle, Henri Sivonen

On 31.01.2012 17:13, Mark Finkle wrote:
> On 01/31/2012 09:46 AM, Henri Sivonen wrote:
>>> Fennec will use the following UA for phones:
>>> "Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131
>>> Firefox/12.0a1 Mobile/12.0a1"
>>
>> Why is there a "/12.0a1" part after "Mobile"?
>
> The HTTP 1.1 spec requires a "/" in product tokens

I'm pretty sure it doesn't. Anyway, as far as I can tell there's no real
need to make this a product token in the first place.

dao

Jonathan Kew

unread,

Jan 31, 2012, 11:35:50 AM1/31/12

to dev-pl...@lists.mozilla.org

According to http://www.w3.org/Protocols/rfc2616/rfc2616-sec3.html#sec3.8,

product = token ["/" product-version]

which means the "/" should be optional.

I believe some existing sniffing code expects to see a "/" after certain tokens, and won't recognize them without it, but does this consideration apply to "Mobile"?

JK

Dao

unread,

Jan 31, 2012, 11:38:55 AM1/31/12

to

On 31.01.2012 17:35, Jonathan Kew wrote:
> On 31 Jan 2012, at 16:23, Dao wrote:
>
>> On 31.01.2012 17:13, Mark Finkle wrote:
>>> On 01/31/2012 09:46 AM, Henri Sivonen wrote:
>>>> Why is there a "/12.0a1" part after "Mobile"?
>>>
>>> The HTTP 1.1 spec requires a "/" in product tokens
>>
>> I'm pretty sure it doesn't. Anyway, as far as I can tell there's no real need to make this a product token in the first place.
>
> According to http://www.w3.org/Protocols/rfc2616/rfc2616-sec3.html#sec3.8,
>
> product = token ["/" product-version]
>
> which means the "/" should be optional.

Yep.

> I believe some existing sniffing code expects to see a "/" after certain tokens, and won't recognize them without it, but does this consideration apply to "Mobile"?

It doesn't. It's a problem for "Gecko" because Webkit has "Gecko" in its
UA string.

dao

Asa Dotzler

unread,

Jan 31, 2012, 11:52:27 AM1/31/12

to

On 1/31/2012 7:02 AM, Ben Bucksch wrote:
> A consent on this newsgroup *is* the project decision. Gerv also is the
> owner of the UA string spec.

Really? I thought Module Owners, rather than newsgroup consensus, were
responsible for final decisions. When did we move to the "if some folks
in a newsgroup agree, they trump module owners" governance system?

- A

Mark Finkle

unread,

Jan 31, 2012, 11:53:41 AM1/31/12

to Dao, Henri Sivonen

If I interpret this correctly, you are right about the spec:
http://www.w3.org/Protocols/rfc2616/rfc2616-sec3.html#sec3.8

The other reason we went with a product token was a higher ranking in
John Jensen's data.

Mark Finkle

unread,

Jan 31, 2012, 11:55:14 AM1/31/12

to Jonathan Kew

On 01/31/2012 11:35 AM, Jonathan Kew wrote:
>
> According to http://www.w3.org/Protocols/rfc2616/rfc2616-sec3.html#sec3.8,
>
> product = token ["/" product-version]
>
> which means the "/" should be optional.
>
> I believe some existing sniffing code expects to see a "/" after certain tokens, and won't recognize them without it, but does this consideration apply to "Mobile"?

The data collected by Jensen shows the using a "/" results in a higher
"content like android stock browser" ranking.

Dao

unread,

Jan 31, 2012, 12:02:54 PM1/31/12

to Asa Dotzler

Really, but your interpretation of Ben's statement is a bit off. Module
Owners should decide if there's no consensus. They may also override
consensus, but then I'd expect them to already voice their opinion at
the time the consensus is being reached.

Also, technically, the Module Owner would be Christian Biesinger in this
case ;)

Asa Dotzler

unread,

Jan 31, 2012, 12:11:20 PM1/31/12

to

On 1/31/2012 9:02 AM, Dao wrote:
> On 31.01.2012 17:52, Asa Dotzler wrote:
>> On 1/31/2012 7:02 AM, Ben Bucksch wrote:
>>> A consent on this newsgroup *is* the project decision. Gerv also is the
>>> owner of the UA string spec.
>>
>> Really? I thought Module Owners, rather than newsgroup consensus, were
>> responsible for final decisions. When did we move to the "if some folks
>> in a newsgroup agree, they trump module owners" governance system?
>
> Really, but your interpretation of Ben's statement is a bit off. Module
> Owners should decide if there's no consensus. They may also override
> consensus, but then I'd expect them to already voice their opinion at
> the time the consensus is being reached.

This is completely wrong. The Module Owner is not just a tie-breaker or
over-rider. The Module Owner is *the* authority. If the Module Owner
delegates decision making to "some people in a newsgroup" and says that
their consensus is good enough, so be it. Unless that has happened, some
coming together of interested parties in a newsgroup has no actual
authority what so ever.

> Also, technically, the Module Owner would be Christian Biesinger in this
> case ;)

I don't agree with this. Christian has nothing to do with Fennec.

- A

Boris Zbarsky

unread,

Jan 31, 2012, 12:14:31 PM1/31/12

to

The general module owner for UA string stuff has been Gerv for a good
long time now, yes?

(I know you think this should be in the purview of the Fennec module
owners, not Gerv, but you could make the exact same argument for the set
of supported CSS properties that Fennec ships. Maybe you think that
should be up to the Fennec module owners too, of course.)

-Boris

Dao

unread,

Jan 31, 2012, 12:18:03 PM1/31/12

to Mark Finkle

I find it hard to say this based on the spreadsheet, since it lists lots
combinations with "Mobile/12.0a1" while "Mobile;" is somewhere in the
middle with about 1.3% less compatibility than the string with the best
rating (not the one you appear to have decided for). What I noticed is
that none of the higher rated strings contain "rv:12.0a1", so I suspect
this plays a larger role.

dao

Asa Dotzler

unread,

Jan 31, 2012, 12:20:01 PM1/31/12

to

Yes, I think this should be in the purview of the Fennec Module Owners.
And I believe that they should also be in control of which Gecko
features they ship or don't ship given that they're the experts in what
their platform can well support.

But I'm actually more concerned about Ben's belief that consensus from a
collection of interested people in a newsgroup is how decisions are made
on this project and that the role of the module owner is to abide by
newsgroup consensus or over-rule it. That's not my understanding of our
governance structure.

- A

Daniel Cater

unread,

Jan 31, 2012, 12:29:31 PM1/31/12

to Jonathan Kew

That's odd, given that stock Android doesn't have a version in its "Mobile" product token.

Are you sure that your interpretation of the results is correct?

Dao

unread,

Jan 31, 2012, 12:29:46 PM1/31/12

to

On 31.01.2012 18:11, Asa Dotzler wrote:
> On 1/31/2012 9:02 AM, Dao wrote:
>> On 31.01.2012 17:52, Asa Dotzler wrote:
>>> On 1/31/2012 7:02 AM, Ben Bucksch wrote:
>>>> A consent on this newsgroup *is* the project decision. Gerv also is the
>>>> owner of the UA string spec.
>>>
>>> Really? I thought Module Owners, rather than newsgroup consensus, were
>>> responsible for final decisions. When did we move to the "if some folks
>>> in a newsgroup agree, they trump module owners" governance system?
>>
>> Really, but your interpretation of Ben's statement is a bit off. Module
>> Owners should decide if there's no consensus. They may also override
>> consensus, but then I'd expect them to already voice their opinion at
>> the time the consensus is being reached.
>
> This is completely wrong. The Module Owner is not just a tie-breaker or
> over-rider. The Module Owner is *the* authority. If the Module Owner
> delegates decision making to "some people in a newsgroup" and says that
> their consensus is good enough, so be it. Unless that has happened, some
> coming together of interested parties in a newsgroup has no actual
> authority what so ever.

This weren't "some people in a newsgroup". This was the platform
newsgroup with every community member invited to provide feedback. I
interpreted the silence here as decision delegating. Again, the Module
Owner has every right to reject that decision, but I would consider it
good behavior to not remain completely silent up to that point.

>> Also, technically, the Module Owner would be Christian Biesinger in this
>> case ;)
>
> I don't agree with this. Christian has nothing to do with Fennec.

Similarly, Christian has nothing to do with desktop Firefox. Or
Thunderbird, XULRunner, you name it. So Christian can't actually make
any decisions that affect some products. Poor Christian.

dao

Boris Zbarsky

unread,

Jan 31, 2012, 12:32:31 PM1/31/12

to

On 1/31/12 12:20 PM, Asa Dotzler wrote:
> Yes, I think this should be in the purview of the Fennec Module Owners.

Then why are we bothering with any module owners at all, except the
Firefox and Fennec ones, if they aren't supposed to make decisions about
their modules?

> And I believe that they should also be in control of which Gecko
> features they ship or don't ship given that they're the experts in what
> their platform can well support.

I guess we'll just have to disagree.

> But I'm actually more concerned about Ben's belief that consensus from a
> collection of interested people in a newsgroup is how decisions are made
> on this project

I think you're completely misreading what Ben said. But whatever.

-Boris

Daniel Cater

unread,

Jan 31, 2012, 12:34:57 PM1/31/12

to

If you will be sticking with the date as the version part of the Gecko product token, could you at least freeze it to the same string that desktop has? The reasoning for that was to stop websites making decisions based on the date instead of the Gecko version (and often getting it wrong).

Dao

unread,

Jan 31, 2012, 12:36:19 PM1/31/12

to

On 31.01.2012 18:34, Daniel Cater wrote:
> If you will be sticking with the date as the version part of the Gecko product token, could you at least freeze it to the same string that desktop has? The reasoning for that was to stop websites making decisions based on the date instead of the Gecko version (and often getting it wrong).

The date is frozen in release versions but not in nightlies. This is
true for both desktop and mobile.

dao

Daniel Cater

unread,

Jan 31, 2012, 12:43:30 PM1/31/12

to

I realise that for desktop, but as this thread used a nightly build for the example in the first post, I wasn't sure if that applied to mobile as well or not. If the mobile release builds also use the frozen string then that would be fine. I'm not sure that's the case though.

Robert Kaiser

unread,

Jan 31, 2012, 1:15:35 PM1/31/12

to

Mark Finkle schrieb:

> As you can see, this is a very minimal change from the current UA, and
> does not have some of the changes discussed over the last weeks,
> summarized here:
> https://wiki.mozilla.org/Fennec/User_Agent

*frown*

All that wasted time on trying to discuss a reasonable solution, when
the decision is made in apparent ignorance of it anyhow.

But then, I lobbied for consistency between desktop and mobile not
without reason. Just too bad we need to add *yet another* token with the
same version number when there seems to be some consensus that the Gecko
token will some time in the future repeat this same version as well.

Robert Kaiser

Robert O'Callahan

unread,

Jan 31, 2012, 3:44:03 PM1/31/12

to Dao, dev-pl...@lists.mozilla.org

On Wed, Feb 1, 2012 at 6:29 AM, Dao <d...@design-noir.de> wrote:

> This weren't "some people in a newsgroup". This was the platform newsgroup
> with every community member invited to provide feedback. I interpreted the
> silence here as decision delegating. Again, the Module Owner has every
> right to reject that decision, but I would consider it good behavior to not
> remain completely silent up to that point.

For what it's worth, I totally agree and think this was badly done.

Rob
--
"If we claim to be without sin, we deceive ourselves and the truth is not
in us. If we confess our sins, he is faithful and just and will forgive us
our sins and purify us from all unrighteousness. If we claim we have not
sinned, we make him out to be a liar and his word is not in us." [1 John
1:8-10]

Gervase Markham

unread,

Jan 31, 2012, 4:18:29 PM1/31/12

to Asa Dotzler

On 31/01/12 17:20, Asa Dotzler wrote:
> But I'm actually more concerned about Ben's belief that consensus from a
> collection of interested people in a newsgroup is how decisions are made
> on this project and that the role of the module owner is to abide by
> newsgroup consensus or over-rule it. That's not my understanding of our
> governance structure.

I would politely suggest that, in this case, Ben's opinions are a rat hole,

Perhaps you could comment on whether you think that it's OK for us to
develop a consensus through discussion with all stakeholders, agree it
as the final decision in a meeting which involved the mobile team (last
Thursday, 5pm GMT, Mobile Evangelism call), post to the newsgroup with
that information with Mark Finkle's blessing... and then for it all to
be thrown over at the 11th hour?

Gerv

Mark Finkle

unread,

Jan 31, 2012, 4:20:15 PM1/31/12

to rob...@ocallahan.org, Dao

On 01/31/2012 03:44 PM, Robert O'Callahan wrote:
> On Wed, Feb 1, 2012 at 6:29 AM, Dao<d...@design-noir.de> wrote:
>
>> This weren't "some people in a newsgroup". This was the platform newsgroup
>> with every community member invited to provide feedback. I interpreted the
>> silence here as decision delegating. Again, the Module Owner has every
>> right to reject that decision, but I would consider it good behavior to not
>> remain completely silent up to that point.
>
>
> For what it's worth, I totally agree and think this was badly done.

The previous thread in which Gerv creates a proposed UA string started
on Jan 27. He managed to reach some form of consensus in the newsgroups,
which is no easy task. Unfortunately I was not able to get the same
consensus in the Mobile team.

After working over the weekend with the team, we were at an impasse. The
simplest thing to do was to revert the UA to a very minimal change that
might get our desired goal: Make it easier for web sites to send Fennec
mobile content.

Our goal was never to re-vamp the UA string. The UA string is not the
silver bullet to Fennec's problem of being served crappy mobile content.

I posted this fallback response on Jan 31. I rejected the re-vamped UA
string. I did not want to fight that battle. I do want to get some means
of allowing web sites to send Fennec decent mobile content.

The platform developers and community are still able to work on
re-vamping the UA. That is not my goal, but mobile will accept changes
to the core UA, just like we did in Firefox 4 timeframe
(http://hacks.mozilla.org/2010/09/final-user-agent-string-for-firefox-4/).

Did I handle this badly? The response makes me think so. Would I do it
again? Probably.

Gervase Markham

unread,

Jan 31, 2012, 4:21:32 PM1/31/12

to Mark Finkle, Henri Sivonen

On 31/01/12 16:13, Mark Finkle wrote:
> The HTTP 1.1 spec requires a "/" in product tokens

Henri told me yesterday that this was not true - what is disallowed is
_ending_ with a /. And, regardless of whether it's true or not, it's
worth noting that other browsers use a bare "Mobile" token.

>> You could drop the "Linux armv7l" part on mobile even if the analogous
>> part is staying on desktop. Especially if the "Mobile"/"Tablet" token
>> was placed in that slot as seemed to be the consensus earlier.
>
> We could, and we might - but it's not pressing right now.

So when do you plan to revisit the question of what Fennec's UA should be?

Gerv

Gervase Markham

unread,

Jan 31, 2012, 4:26:21 PM1/31/12

to

On 31/01/12 16:55, Mark Finkle wrote:
> The data collected by Jensen shows the using a "/" results in a higher
> "content like android stock browser" ranking.

That is not proven. There were no two user agents submitted which
differed only in this way, so you can't draw this conclusion. (If I'm
wrong, show me how you are drawing it.)

There was "Firefox/X Mobile/X' vs. 'Firefox Mobile/X" but, as my
analysis of the data shows:

http://groups.google.com/group/mozilla.dev.platform/browse_thread/thread/8fe2f393673b4fa0/b127e416d3f9e71f

the effect of this is almost lost in the statistical noise.

'Firefox/X Mobile/X' vs. 'Firefox Mobile/X'

This has only a small effect (0.0033, 0.0036, 0.0037), with 'Firefox/X
Mobile/X' being better - perhaps because the latter means that an
attempt to parse a Firefox version number will fail, or perhaps because
people are looking for "Firefox/".

Gerv

Gervase Markham

unread,

Jan 31, 2012, 4:27:52 PM1/31/12

to

On 31/01/12 16:53, Mark Finkle wrote:
> The other reason we went with a product token was a higher ranking in
> John Jensen's data.

Again, this is not proven, because the data does not contain two user
agents which differ only in this respect. All of the blassey-submitted
strings had "Mobile" as a product token. My strings did not, but they
are too different from his to draw comparisons between one of mine and
one of his.

Gerv

Asa Dotzler

unread,

Jan 31, 2012, 4:38:12 PM1/31/12

to

I didn't follow this that closely but this is what it looked like to me:

I saw a lot of interested parties discussing options here. I saw several
proposals and lots of debate. I saw two or three people almost come to
some agreement around one particular proposal but with plenty of dissent
still being communicated. Then I saw a Fennec lead say "hey, thanks for
all the help, we settled on something pretty close to what was proposed
here."

- A

Mark Finkle

unread,

Jan 31, 2012, 4:38:48 PM1/31/12

to Gervase Markham, Henri Sivonen

On 01/31/2012 04:21 PM, Gervase Markham wrote:
> On 31/01/12 16:13, Mark Finkle wrote:
>> The HTTP 1.1 spec requires a "/" in product tokens
>
> Henri told me yesterday that this was not true - what is disallowed is
> _ending_ with a /. And, regardless of whether it's true or not, it's
> worth noting that other browsers use a bare "Mobile" token.

Yeah. The incorrect reading of the spec is really the only reason I had
for using the "Mobile/VERSION" over just the bare "Mobile"

"Mobile/VERSION" has a tiny jump in Jensen ranking, maybe from the
influence of the "Mobile/VERSION" used in mobile Safari?

>>> You could drop the "Linux armv7l" part on mobile even if the analogous
>>> part is staying on desktop. Especially if the "Mobile"/"Tablet" token
>>> was placed in that slot as seemed to be the consensus earlier.
>>
>> We could, and we might - but it's not pressing right now.
>
> So when do you plan to revisit the question of what Fennec's UA should be?

Honestly, _I_ am not eager to revisit the UA. My primary goal here was
to give UA sniffers a way to increase the probability of sending Fennec
mobile content. Adding the "Mobile" token might be enough.

So for me the question becomes "When do we revisit the question of what
Mozilla's core UA should be?"

Robert O'Callahan

unread,

Jan 31, 2012, 4:57:53 PM1/31/12

to Mark Finkle, Dao, dev-pl...@lists.mozilla.org

On Wed, Feb 1, 2012 at 10:20 AM, Mark Finkle <mark....@gmail.com> wrote:

> The previous thread in which Gerv creates a proposed UA string started on
> Jan 27. He managed to reach some form of consensus in the newsgroups, which
> is no easy task. Unfortunately I was not able to get the same consensus in
> the Mobile team.
>

If there are people on the Mobile team who aren't willing to discuss their
concerns in public threads, but are willing to veto the results of those
public threads, maybe that's the problem.

Ben Bucksch

unread,

Jan 31, 2012, 5:11:04 PM1/31/12

to

On 31.01.2012 18:32, Boris Zbarsky wrote:
> On 1/31/12 12:20 PM, Asa Dotzler wrote:
>> Yes, I think this should be in the purview of the Fennec Module Owners.
>
> Then why are we bothering with any module owners at all, except the
> Firefox and Fennec ones, if they aren't supposed to make decisions
> about their modules?

Of course, you're right. Firefox module owners only own the UI (browser/
), nothing else. Ditto Fennec. Anything about network is netwerk/, and
Biesi is the owner.

As for the UA string, Gerv has always been the owner of that spec, as I
said. Clearly, he had a very different idea, and apparently, he was not
involved in the decision either. So, Asa, you are violating the very
government structure you are referring to.

Unless of course the structure is that you're the big boss and overrule
everybody apart from Mitchell. I do believe that this is a Free Software
project and must reach decisions openly and with the contributors, not
ambush them with secret decisions like that. We had too much of that
recently.

Ben

Brad Lassey

unread,

Jan 31, 2012, 6:26:55 PM1/31/12

to mozilla.de...@googlegroups.com, Dao, Mark Finkle, dev-pl...@lists.mozilla.org, rob...@ocallahan.org

Roc, where did anyone say that? Discussions about the UA string were happening in many different places, including a bug before this series of newsgroup discussions started happening. I believe bugzilla would still count as public.

As Gerv and the team discussed on the call he referenced in his previous post on the subject, his goals and the goals of the Fennec product are not aligned. For the product we simply want a UA which gets us useful content with the simplest change. I don't think you can get much simpler than what we arrived at (substituting Mobile for Fennec). That change is entirely with in the purview of Fennec module owner, which is Finkle. I also might add that after all this discussion we arrived at the same decision we had in the bug before the newsgroup discussion started.

The relatively simple goal of the bug was hijacked by these newsgroup threads to accomplish other competing goals, such as fighting fingerprinting and helping out b2g whenever it comes along. Since we've identified those as non-goals, those discussions are really tangential to the actual change being discussed. Anyone who cares significantly about those goals should file the appropriate bugs to make those changes across the platform, not just within the Fennec product.

Brad Lassey

unread,

Jan 31, 2012, 6:26:55 PM1/31/12

to Mark Finkle, Dao, dev-pl...@lists.mozilla.org, rob...@ocallahan.org

Ben Bucksch

unread,

Jan 31, 2012, 7:03:03 PM1/31/12

to

On 01.02.2012 00:26, Brad Lassey wrote:
> his goals and the goals of the Fennec product are not aligned. For the product we simply want a UA which gets us useful content with the simplest change.

Exactly. Gerv has a longer-term view. You just showed very well what is
going wrong here.

We're not in it for the quick buck. But even those who are, and who
started with 0 market share, namely Google Android browser, made a
decision more brave than that.

> Discussions about the UA string were happening in many different places, including a bug before this series of newsgroup discussions started happening. I believe bugzilla would still count as public.

> I also might add that after all this discussion we arrived at the same decision we had in the bug before the newsgroup discussion started.

Which translates to "we entirely ignored the discussion here".

> That change is entirely with in the purview of Fennec module owner

No, that is not correct.

Ben

Jonas Sicking

unread,

Jan 31, 2012, 7:45:53 PM1/31/12

to Boris Zbarsky, dev-pl...@lists.mozilla.org

On Tue, Jan 31, 2012 at 9:32 AM, Boris Zbarsky <bzba...@mit.edu> wrote:
> On 1/31/12 12:20 PM, Asa Dotzler wrote:
>>
>> Yes, I think this should be in the purview of the Fennec Module Owners.
>
> Then why are we bothering with any module owners at all, except the Firefox
> and Fennec ones, if they aren't supposed to make decisions about their
> modules?
>
>> And I believe that they should also be in control of which Gecko
>> features they ship or don't ship given that they're the experts in what
>> their platform can well support.
>
> I guess we'll just have to disagree.

I agree with Boris here.

I don't think that simply allowing the Firefox and Fennec module
owners override any decision made by other module owners, because the
code is shipping in those products, accurately reflects the intent of
the module system.

This is *especiallly* true now that we have multiple products since it
means that we'll want to coordinate between those products to make
sure that we create a coherent platform for our users and web
developers.

For example if the B2G team decided that they wanted some DOM API to
work slightly differently then I think think the DOM module owners
should have the final say on that matter. Having consistent DOM APIs
is important to web developers and so B2G going it's own way could
harm Fennec adoption if web pages don't work in both.

I don't have a great answer about what to do if module owners for a
product disagrees with the module owners of a component. I guess if
the two can't come to an agreement it rolls up to the top module owner
which is Brendan. But I certainly don't think the answer is for the
product module owner to simply call the shots and override anyone
else.

/ Jonas

Matt Brubeck

unread,

Jan 31, 2012, 7:47:39 PM1/31/12

to

On 01/31/2012 04:03 PM, Ben Bucksch wrote:
> We're not in it for the quick buck. But even those who are, and who
> started with 0 market share, namely Google Android browser, made a
> decision more brave than that.

Could you clarify this? For reference, the user agent of the Android
Browser on an Android 2.3 device is:

Mozilla/5.0 (Linux; Android 2.3.4; en-us; T-Mobile G2 Build/GRJ22)
AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1

That seems to be the most pragmatic, legacy-focused string possible. It
contains, among other things, the names of two operating systems, two
companies, three rendering engines, two competing browsers, and six
version identifiers.

Daniel Cater

unread,

Jan 31, 2012, 8:40:15 PM1/31/12

to

On Tuesday, 31 January 2012 21:38:12 UTC, Asa Dotzler wrote:
> Then I saw a Fennec lead say "hey, thanks for
> all the help, we settled on something pretty close to what was proposed
> here."
>
> - A

Mozilla/5.0 (Android; Linux armv7l; rv:12.0a1) Gecko/20120131 Firefox/12.0a1 Mobile/12.0a1

and

Mozilla/5.0 (Android; Mobile; rv:12.0) Gecko/12.0 Firefox/12.0

are not pretty close. They differ in 5 important ways:

Compared to the string from the wiki, the string in the patch had:

The presence of "Linux" in the comments section.
The presence of "armv7l" in the comments section.
The absence of "Mobile" in the comments section.
The addition of a "Mobile/version" product token.
Gecko/date instead of Gecko/version.

Asa Dotzler

unread,

Jan 31, 2012, 8:50:23 PM1/31/12

to

On 1/31/2012 4:45 PM, Jonas Sicking wrote:

> For example if the B2G team decided that they wanted some DOM API to
> work slightly differently then I think think the DOM module owners
> should have the final say on that matter. Having consistent DOM APIs
> is important to web developers and so B2G going it's own way could
> harm Fennec adoption if web pages don't work in both.

I agree that we should not ship competing versions of a standard or
standards-track API in different Firefox products except by the accident
of ship schedules meaning more or less completeness in one or the other
product. Shipping competing versions of DOM APIs gets no one anything of
benefit.

But that isn't the end of the story nor is that what we're talking about
here.

I don't believe that the absentee Networking module owner owns the User
Agent string for Mozilla Products like Fennec or B2G, (or Desktop,
though I think we've mostly been OK with consensus changes there -- not
so much in the http headers changes where I've challenged and had
reverted because broke the web for very little or no user benefit.)

I think Necko should be a part of the discussion, but I don't think that
it can be solely their discretion. Nor do I think that Gerv is the final
say over what Fennec ships as their user agent. I appreciate the work
he's put in here and everyone else, but what the Fennec team are trying
to accomplish and what Gerv and Necko folks are trying to accomplish
obviously aren't aligned.

BTW, to get back to your original example, here's my take. If B2G
decided to not ship a particular Desktop-focused DOM API because no B2G
target device even had that hardware capability or Firefox Desktop
decided not to ship a phone-specific API because it offered no use
benefit and only security surface, I think that should absolutely be the
responsibility of the B2G and Desktop leadership and not the DOM
leadership. I'd say the same for Mobile. If Mobile decided that they
couldn't possibly make a good WebGL experience, for example, because of
the characteristics of their product or the devices their product runs
on, I think it's their call if they want to disable WebGL and not the
call of the Graphics team.

I'm not asserting that is any kind of shared understanding today, but
that's what I think.

- A

unread,

Feb 1, 2012, 3:03:07 AM2/1/12

to Asa Dotzler

He's not as present as others who are paid to be present 40 hours a
week. I don't think he's absent, though. The last message I read from
him was:
"Re: Networking Team Meeting: Tuesday, January 24, 10am Pacific. My
apologies, I'm unable to make it to this week's meeting."

Anyway, as Callek said, if you think Module Owners don't execute their
role, you need to bring that up rather than bypassing them.

dao

Asa Dotzler

unread,

Feb 1, 2012, 3:04:48 AM2/1/12

to

And just to be clear, I don't mean that Biesi is absent on important
Necko issues but that he's not a part of this discussion (which I don't
consider an important Necko issue.)

- A

Dao

unread,

OK, then I retract that.

Ben

Ben Bucksch

unread,

Feb 1, 2012, 9:04:00 AM2/1/12

to

On 01.02.2012 04:00, Jonas Sicking wrote:

> If people can't agree for whatever reason, such as goals not being
> align, then that needs to be explicitly resolved through discussion.
> Not through one side simply forging ahead.

+1

Ben Bucksch

Justin Dolske

unread,

Feb 3, 2012, 4:05:12 AM2/3/12

to

On 1/31/12 4:47 PM, Matt Brubeck wrote:

> On 01/31/2012 04:03 PM, Ben Bucksch wrote:
>> We're not in it for the quick buck. But even those who are, and who
>> started with 0 market share, namely Google Android browser, made a
>> decision more brave than that.
>

> Could you clarify this? For reference, the user agent of the Android
> Browser on an Android 2.3 device is:
>
> Mozilla/5.0 (Linux; Android 2.3.4; en-us; T-Mobile G2 Build/GRJ22)
> AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1
>
> That seems to be the most pragmatic, legacy-focused string possible. It
> contains, among other things, the names of two operating systems, two
> companies, three rendering engines, two competing browsers, and six
> version identifiers.

Haha, I'd love to tweet either the UA or your description. But neither
will fit (hint!).

[Ok, techically the bare UA will fit, but with only 3 characters to
spare it's challenging to say anything about it.]

Justin

Fennec User Agent - The End

Mark Finkle

Henri Sivonen

Dao

Ben Bucksch

Mark Finkle

Dao

Jonathan Kew

Dao

Asa Dotzler

Mark Finkle

Mark Finkle

Dao

Asa Dotzler

Boris Zbarsky

Dao

Asa Dotzler

Daniel Cater

Dao

Boris Zbarsky

Daniel Cater

Dao

Daniel Cater

Robert Kaiser

Robert O'Callahan

Gervase Markham

Mark Finkle

Gervase Markham

Gervase Markham

Gervase Markham

Asa Dotzler

Mark Finkle

Robert O'Callahan

Ben Bucksch

Brad Lassey

Brad Lassey

Ben Bucksch

Jonas Sicking

Matt Brubeck

Daniel Cater

Asa Dotzler

Justin Wood (Callek)

Jonas Sicking

Brad Lassey

Brad Lassey

Asa Dotzler

Brad Lassey

Brad Lassey

Dao

Asa Dotzler

Dao

Asa Dotzler

Asa Dotzler

Ben Bucksch

Ben Bucksch

Ben Bucksch

Ben Bucksch

Ben Bucksch

Ben Bucksch

Boris Zbarsky

Dave Townsend

Justin Wood (Callek)

Brian Smith

Henri Sivonen

Ben Bucksch

Robert Kaiser

Justin Dolske