Testing new UA string for Firefox

[kim....@safe-mail.net]

I have been testing a new Firefox User-Agent header. Using this reference:

https://developer.mozilla.org/en-US/docs/Web/HTTP/Gecko_user_agent_string_reference

I have selected:

Gecko/geckoversion Firefox/firefoxversion (platform)

An example:

Gecko/31.0 Firefox/31.0 (Windows NT 6.1)

I hoped to find some broken sites, so I ran a test over Alexa's site list, but found no breakage. I also manually tested many sites that use javascript parsing of UA string, but found no breakage either. Is it possible all websites have been fixed to not do broken UA detection?

Can somebody please send me a site that breaks with the above proposed UA string?

[Gervase Markham]

On 07/03/15 20:11, kim....@Safe-mail.net wrote:
> I have been testing a new Firefox User-Agent header. Using this
> reference:
>
> https://developer.mozilla.org/en-US/docs/Web/HTTP/Gecko_user_agent_string_reference

Why? What do you hope to achieve? :-)

> I have selected:
>
> Gecko/geckoversion Firefox/firefoxversion (platform)
>
> An example:
>
> Gecko/31.0 Firefox/31.0 (Windows NT 6.1)

Out of interest, why did you choose those elements, and why in that order?

> I hoped to find some broken sites, so I ran a test over Alexa's site
> list, but found no breakage.

Which site list - top 10? top 100? top 1000?

How much of the sites did you visit - just the front page?

How did you determine "no breakage" - manual inspection?

What about mobile?

> I also manually tested many sites that
> use javascript parsing of UA string, but found no breakage either. Is
> it possible all websites have been fixed to not do broken UA
> detection?

No :-)

We've done quite a bit of compatibility work on UAs over the past couple
of years, working out the best UA for Firefox for Android and for
Firefox OS. And, no doubt, we soon have to do the same for Servo.
Removing Mozilla/5.0 actually has little compatibility impact, so we
could do that (although it would be odd for Mozilla to remove "Mozilla"
from their UA). Every change results in non-zero breakage, so it's
always a trade-off.

Gerv

[kim....@safe-mail.net]

> Why? What do you hope to achieve? :-)

For the same reason Mozilla developed Australis :-)

> > Gecko/31.0 Firefox/31.0 (Windows NT 6.1)
> Out of interest, why did you choose those elements, and why in that order?

Rendering engine > Application name > Platform. Order of importance.

> Which site list - top 10? top 100? top 1000?

top 10000

> How much of the sites did you visit - just the front page?

Yes.

> How did you determine "no breakage" - manual inspection?

Scripted run of Firefox with modified UA, and normal UA, both at the same time. Both having installed ublock for adblocking, to remove changing advertisements. Saving the page as image from both, then diffing the image. Manually inspecting images with large diffs.

> What about mobile?
No testing for mobile, mobile can keep its current UA.

Can I please get an example of a webpage that breaks as a result of UA header being "Gecko/31.0 Firefox/31.0 (Windows NT 6.1)" ?

[Gervase Markham]

On 09/03/15 14:22, kim....@Safe-mail.net wrote:
>>> Gecko/31.0 Firefox/31.0 (Windows NT 6.1)
>> Out of interest, why did you choose those elements, and why in that
>> order?
>
> Rendering engine > Application name > Platform. Order of importance.

Would it not make more sense to keep them in the same order as now, to
improve compatibility?

> Can I please get an example of a webpage that breaks as a result of
> UA header being "Gecko/31.0 Firefox/31.0 (Windows NT 6.1)" ?

I'm afraid it doesn't work like that. We don't have to disprove your
ideas. :-) You haven't given us any gains that we could expect to see
from the change, so it's not worth evaluating it.

Gerv

[kim....@safe-mail.net]

>> Rendering engine > Application name > Platform. Order of importance.
> Would it not make more sense to keep them in the same order as now, to
> improve compatibility?

I have no emotional attachment to the order, just came up with that one as a personal preference.

>> Can I please get an example of a webpage that breaks as a result of
>> UA header being "Gecko/31.0 Firefox/31.0 (Windows NT 6.1)" ?
> I'm afraid it doesn't work like that. We don't have to disprove your
> ideas. :-) You haven't given us any gains that we could expect to see
> from the change, so it's not worth evaluating it.

I am mainly asking for an example to see if my testing method is valid or not. When I setup my own webpage that specifically breaks when it detects the UA as being non-IE, then I obviusly get a nice diff image. But I found no such breakage in top 10000 (!) webpages, which is hard for me to believe, hence I need an example of a page that breaks.

Reasoning for change of UA:
1) Decreasing the size of HTTP headers, so that all the headers combined fall below various thresholds (e.g. fitting into 1 packet, or fitting below certain size for proxies) can improve internet connectivity as a whole, especially when not using HTTP/2.

2) Removing things from UA, for all users of a browser, improves users' privacy by limiting identifiable fields.

[Ehsan Akhgari]

On 2015-03-13 8:15 PM, kim....@Safe-mail.net wrote:
>>> Can I please get an example of a webpage that breaks as a result of
>>> UA header being "Gecko/31.0 Firefox/31.0 (Windows NT 6.1)" ?
>> I'm afraid it doesn't work like that. We don't have to disprove your
>> ideas. :-) You haven't given us any gains that we could expect to see
>> from the change, so it's not worth evaluating it.
> I am mainly asking for an example to see if my testing method is valid or not. When I setup my own webpage that specifically breaks when it detects the UA as being non-IE, then I obviusly get a nice diff image. But I found no such breakage in top 10000 (!) webpages, which is hard for me to believe, hence I need an example of a page that breaks.

For this purpose, try <http://whatsmyuseragent.com/>. :-)

Cheers,
Ehsan

[Gavin Sharp]

On Fri, Mar 13, 2015 at 5:15 PM, <kim....@safe-mail.net> wrote:
> I am mainly asking for an example to see if my testing method is valid or not. When I setup my own webpage that specifically breaks when it detects the UA as being non-IE, then I obviusly get a nice diff image. But I found no such breakage in top 10000 (!) webpages, which is hard for me to believe, hence I need an example of a page that breaks.

If you are just loading the front page of the top 10000 web sites, you
are only scratching the surface of potential compatibility issues. For
example, just loading https://www.facebook.com is very different than
logging in to Facebook, chatting with your friends, posting/browsing
pictures, creating events, etc. A similar situation exists with most
sites that require logins.

Gavin

[Gervase Markham]

On 14/03/15 00:15, kim....@Safe-mail.net wrote:
> Reasoning for change of UA: 1) Decreasing the size of HTTP headers,
> so that all the headers combined fall below various thresholds (e.g.
> fitting into 1 packet, or fitting below certain size for proxies) can
> improve internet connectivity as a whole, especially when not using
> HTTP/2.

Theoretically it can, but previously we have decided that the tiny
number of bytes saved by messing with the UA string is generally not
worth the compatibility hit.

> 2) Removing things from UA, for all users of a browser, improves
> users' privacy by limiting identifiable fields.

You are sending:

Gecko/31.0 Firefox/31.0 (Windows NT 6.1)

Firefox currently sends (for me):
Mozilla/5.0 (X11; Linux i686; rv:38.0) Gecko/20100101 Firefox/38.0

Mozilla/5.0 is a fixed string.
20100101 is a fixed string.

Your string includes OS info; so does the current one.

So in what way is privacy improved with your string?

Gerv