Avoid having to re-parse URLs after deserializing them. This impacts both our IPC and our disk cache storage code. We currently serialize URLs as Strings, which means that we have the re-parse them upon deserialization.
Created attachment 278807 [details] Patch
Comment on attachment 278807 [details] Patch One encoder is better than the two we had before. r=me for correctness, but is this really faster? Also, why are all the member variables signed integers? I think they should be unsigned.
(In reply to comment #2) > Comment on attachment 278807 [details] > Patch > > One encoder is better than the two we had before. r=me for correctness, but > is this really faster? > Also, why are all the member variables signed integers? I think they should > be unsigned. Do you mind if I change the type of those members in a follow up?
Comment on attachment 278807 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=278807&action=review > Source/WebCore/ChangeLog:9 > + Avoid having to re-parse URLs after deserializing them. This impacts > + both our WK2 IPC and our WK2 network cache storage code. Is this good when handling untrusted URLs from WebContent in UI process?
I agree with Alexey that it in theory would be a problem trusting data from the web process. Worth talking over Anders and Sam at some point.
(In reply to comment #5) > I agree with Alexey that it in theory would be a problem trusting data from > the web process. Worth talking over Anders and Sam at some point. Ok, based on Alexey and Darin's feedback, I think I'll restrict the optimization to the disk cache for now and not the IPC.
Created attachment 278858 [details] Patch
Comment on attachment 278858 [details] Patch Clearing flags on attachment: 278858 Committed r200909: <http://trac.webkit.org/changeset/200909>
All reviewed patches have been landed. Closing bug.