[squid-users] pdf and byteranges

We are experiencing problems (2.4.STABLE1) with fetching PDFs
through squid, typically when using NN4 with the Adobe plugin.

The problem seems to be associated with a content type of
multipart/x-byteranges

Mailing-list matches for the symptoms have produced four matches
http://list.cineca.it/cgi-bin/wa?S2=squid&q=byteranges&s=&f=&a=&b=

#5338 and #5358 are quite old; the first one quotes a fix involving
range_offset_limit, but as I read it, the poster claims it doesn't
work; the second ended with a request for the afflicted user to submit
extended access log output, but then the trail goes cold.

Subsequent reports of problems that seem to be at least superficially
similar seem to have gone unanswered(?)

The actual document body that has been retrieved by the client appears
to start like this (in one specimen case):

 ___
/
--3ba9b9faec

Content-type: application/pdf

Content-range: bytes 162409-176436/176437

 (Exceed_DSE520201_R1.indd)
>>
endobj
xref
0 693
0000000000 65535 f
\___

...and so on

Well, I've tried setting range_offset_limit to -1, and the 'squid -k
check' grumbled that there were no units specified. So I set it to
-1 KB, and now it seems to be happy. I issued the reconfigure...

And on the basis of a few documents that previously couldn't be
viewed correctly, the change _does_ seem to have been effective.

However, I'm worried that this might not be the intended approach(?).

Any other suggestions, please?
Received on Thu Sep 20 2001 - 05:01:47 MDT