diff options
author | Monty Brandenberg <monty@lindenlab.com> | 2012-07-09 11:47:47 -0400 |
---|---|---|
committer | Monty Brandenberg <monty@lindenlab.com> | 2012-07-09 11:47:47 -0400 |
commit | 348db20b92f1f1f85712c5a9a862ef079102bbee (patch) | |
tree | 7e19406334296b0ae3da6ff4b7a77400f31edb79 /indra/llcorehttp/httpresponse.h | |
parent | f5f51d3cda8861b3b3a380cc96aaca98e572c377 (diff) |
SH-3187 Issue smarter 'Range' requests for textures.
First, try to issue ranged GETs that are always at least partially
satisfiable. This will keep Varnish-type caches from simply sending
back 200/full asset responses to unsatisfiable requests. Implement
awareness of Content-Range headers as well. Currently they're not
coming back but they will be someday.
Diffstat (limited to 'indra/llcorehttp/httpresponse.h')
-rw-r--r-- | indra/llcorehttp/httpresponse.h | 6 |
1 files changed, 6 insertions, 0 deletions
diff --git a/indra/llcorehttp/httpresponse.h b/indra/llcorehttp/httpresponse.h index 925cf81586..65e403cec8 100644 --- a/indra/llcorehttp/httpresponse.h +++ b/indra/llcorehttp/httpresponse.h @@ -111,6 +111,12 @@ public: /// If a 'Range:' header was used, these methods are involved /// in setting and returning data about the actual response. + /// If both @offset and @length are returned as 0, we probably + /// didn't get a Content-Range header in the response. This + /// occurs with various Capabilities-based services and the + /// caller is going to have to make assumptions on receipt of + /// a 206 status. The @full value may also be zero in cases of + /// parsing problems or a wild-carded length response. void getRange(unsigned int * offset, unsigned int * length, unsigned int * full) const { *offset = mReplyOffset; |