Patch committed in r59178.

Reopening since I'm recommending reverting that revision. No compliant HTTP server is going to treat a PURGE like GET, as the original code seems to be implying, so I think that case can just be removed. If you want to detect an error you should look at the response code.

avinoamr wrote:

Example of a Varnish HTTP response

Attached:

avinoamr wrote:

I disagree.

Varnish allows you to configure whichever output you wish to receive for a PURGE command, and it also adds information about the purge (HTTP response text attached). I realize this might be non-standard, but I believe that keeping MW's code portable (or easy to integrate with different solutions) is a much more meaningful advantage than enforcing standard usage of other products in the market.

Another approach would be to design and implement a similiar VarnishUpdate class, but this might be an overkill.

What do you think?

I think you're missing the point. The response size limit is a useless paranoid hack for Squid, and is harmful for Varnish, and so should be removed for both. Adding a configuration variable is pointless if the whole code block can be deleted.

avinoamr wrote:

Oh, I misunderstood. Great solution.

Done in r62005.