https://bz.apache.org/SpamAssassin/show_bug.cgi?id=8101

--- Comment #9 from Christer Mjellem Strand <[email protected]> ---
(In reply to Sidney Markowitz from comment #7)
> (In reply to Christer Mjellem Strand from comment #6)
> 
>> https://ml-trk.com/
>> aff_c?offer_id=8822&aff_id=37119&url_id=0_15409_15410_15411_15412_15413&aff_s
>> ub5=email&source=s002
>> 
>> Returns a location header of:
>> 
>> /rd.html?go=https%3A%2F%2Fwww.SingleUkrainianBeauty.com%2Findex.
>> php%2Fpromote%2Fclick%3Faid%3D2354%26oid%3DCP231375%26qpid_offer_id%3DSUB_235
>> 084TSAEI%26qpid_subid%3D37119%26qpid_clickid%3D45_37119_8822_52c5875f9b6d77a8
>> b48d37630ad127b4%26source_tag%3Ds002
> 
> If that is not recognized as a redirect to a local page and handled by
> looping to process it, then that should be fixed within the scope of this
> issue.

This was my reasoning too, and why I didn't open a separate issue. In case it
turns out to be relevant, I have (fairly aggressive) caching enabled, and the
first link hits a cached response, but according to logs doesn't follow the
(cached) relative link. This was not a one-off - I've received more spam since
my last comment involving ml-trk.com, which follows this same pattern.

> However, it is a separate issue that it is not detected as a redirect to the
> urldecoded form of the URL that comes after the ?go=
> 
> If I use curl to get the body of the /rd.html?go=... URL, I see it returns a
> page with Javascript that does a redirect to the urldecoded specified URL.
> The plugin will have no way of recognizing the /rd.html?go= form of the
> redirect unless we add it as a special case when handling ml-trk URLs. I've
> opened bug 8114 for dealing with that, and will consider if it is worth
> doing.

Right, I filed bugs for several other cases (see 8109 and 8111), but I didn't
file one for urldecode. That would certainly be nice to support too (preferably
in a generic fashion), in the spirit of handling all types of obfuscation
through redirection, although I would probably put urldecode at the bottom of
that list, below the other (presumably more easily supported) methods I've
suggested.

> However, it is strange that when I run
> 
> curl -I
> 'https://ml-trk.com/
> aff_c?offer_id=8822&aff_id=37119&url_id=0_15409_15410_15411_15412_15413&aff_s
> ub5=email&source=s002'
> 
> it comes back with a 302 redirect with a location header of the
> SingleUkrainianBeauty URL, not what you say the logs showed.
> 
> I'll try running it through SpamAssassin to see if the way the plugin
> fetches the URL gets different results than curl -I, or whether the ml-trk
> site is changing behavior.

As suggested in the last comment, this has every indication of being a case of
UA sniffing. Use the default Mozilla UA that the plugin uses, and you'll see
what I see:

curl -A "Mozilla/5.0 (Windows NT 10.0; Win64 ;x64) AppleWebKit/537.36 (KHTML,
like Gecko) Chrome/101.0.4951.67 Safari/537.36)" -I
"https://ml-trk.com/aff_c?offer_id=8822&aff_id=37119&url_id=0_15409_15410_15411_15412_15413&aff_sub5=email&source=s002";

This would be covered by bug 8110, which I recently filed.

So if you were able to do something like this:

url_shortener_ua ml-trk.com "curl/1.23.4"

Then that should be a functioning workaround for this particular redirector
(and others with similar behavior, such as t.co, as noted in the other bug
report).

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to