Hi Sasha, On Wed, May 21, 2014 at 05:43:27PM -0600, Sasha Pachev wrote: > > No, it would not concatenate the headers, otherwise it breaks > > set-cookie. For simple headers like above, you'd get the same > > result with modify-header and replace-header. The difference > > is visible when the header contains a comma : > > > > Set-Cookie: A=B; expires=Wed, 21 May 2014 05:11:16 GMT > > Set-Cookie: C=D; expires=Wed, 21 May 2014 05:11:16 GMT > > > > replace-header Set-Cookie (.*) \1;bar would give : > > > > Set-Cookie: A=B; expires=Wed, 21 May 2014 05:11:16 GMT; bar > > Set-Cookie: C=D; expires=Wed, 21 May 2014 05:11:16 GMT; bar > > > > modify-header Set-Cookie (.*) \1;bar would give : > > > > Set-Cookie: A=B; expires=Wed; bar, 21 May 2014 05:11:16 GMT; bar > > Set-Cookie: C=D; expires=Wed; bar, 21 May 2014 05:11:16 GMT; bar > > > > Maybe the names are not the best choices, I don't know. At least I > > tend to remember which one does what this way. > > Good plan. Will do it this way. If I use buffer_replace2() on the > header repeatedly from under http_res_get_intercept_rule() or > http_req_get_intercept_rule() it does not appear that it would cause > any problems - It is already being used from under > http_remove_header2() which is called from under those functions, but > maybe I missed something. Did I? > > And, is buffer_replace2() the best tool for the job, or would you > suggest a different call?
Yes I think it's the appropriate one. Look how the Connection header is processed for example. You might also be interested in looking at how reqrep/rsprep are processed in apply_*_filter* with action ACT_REPLACE. I think it's the closest to what you want to do. Best regards, Willy
