[QGIS-Developer] [Qgis-psc] Request for Change of UserAgent

[Jonathan Moules]

This makes sense, but thinking about it, it doesn't seem plausible.

If I were whitelisting user agents, there are three ways I'd consider 
doing it.

Explicit full string:
* "Mozilla/5.0 QGIS/3.4.3-Madeira" - except of course that'd break next 
time someone updates QGIS by even a point version. So I doubt anyone is 
doing that; too much maintenance. And if they *are* doing this, well 
then they'll have to update their whitelist with the new user agent on 
the new release whether it includes "Mozilla" or not.

* Some sort of regexp looking for that structure (i.e. "Mozilla/5.0 
QGIS/[0-9].[0-9]...") but allow any set of numbers. I can't conceive of 
when this would be the best option or even a "good" option - it's just 
asking for problems with version numbers (what happens when QGIS gets to 
double-digits in any version field for instance?). I don't doubt it may 
solve a particular problem somewhere maybe, but I would hope there 
wouldn't be many using this method (and of course, this is assuming that 
the "Mozilla" is part of the regex).

* Search the user agent for the string "QGIS". I do some things that 
look at user-agents and this is what I do. It's easily the simplest and 
definitely the most fool-proof way to validate a client is at least 
claiming to be QGIS.

So I'd suggest this is probably not something to worry about.
Cheers,
Jonathan


On 2019-02-24 22:44, Nyall Dawson wrote:
> On Sun, 24 Feb 2019 at 04:38, Jonathan Moules
> <jonathan-lists at lightpear.com> wrote:
>> Hi Richard,
>>
>> I don't think without more information it's clear precisely what QGIS
>> should change to help them?
>>
>> Indeed, personally I'm still not sure what the current problem is for
>> them. I get that they have some over-users of their service and need to
>> curtail this (a very tricky problem), but what's their current solution
>> and how does a user-agent change help it?
>>
>> I'd suggest against QGIS using different user-agents for destinations as
>> was suggested earlier in the thread because down that road lies madness
>> (it's a highly opaque thing that /will/ bite people as they try and
>> debug issues).
> My main concern is possible regressions if we flip to a
> "less-standard" user agent. It's highly likely that there's servers
> out there which block requests based on user agent whitelists, so by
> changing the user agent across the whole app we'd potentially be
> blocking access to these servers.
>
> Nyall
>
>> Cheers,
>>
>> Jonathan
>>
>>
>> On 23/02/2019 11:17, Richard Duivenvoorde wrote:
>>> For what I understand, they use some cookie-based trick to help maintain
>>> their QOS (and then to make it possible to throttle a very demanding web
>>> application?).
>>>
>>> This made sense to me. Also because I really do not envy maintainers of
>>> such services: it is hard to keep up such free services (as an example
>>> (see irc log) he mentioned that certain transportation software started
>>> to poll the reverse geocoding every second in every car). I think it's
>>> pretty important for us to have OSM, and be helpfull to them.
>>> But if Firefishy/Grant is willing to give more details (he is in bcc)
>>> that would be great.
>>>
>>> Regards,
>>>
>>> Richard
>>>
>>>
>>> _______________________________________________
>>> QGIS-Developer mailing list
>>> QGIS-Developer at lists.osgeo.org
>>> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>>> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>>
>> _______________________________________________
>> QGIS-Developer mailing list
>> QGIS-Developer at lists.osgeo.org
>> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer