Revert "Buffer HPACK parsing until the end of a header boundary"#26825
Revert "Buffer HPACK parsing until the end of a header boundary"#26825
Conversation
…)" This reverts commit 8bab3e4.
ctiller
added
the
release notes: no
ejona86
left a comment
ejona86
left a comment
There was a problem hiding this comment.
As I just commented (#26787 (comment)) you can allow some overbuffering, but not "excessive" overbuffering (tehe) with this approach. I think I did an audit of the implementations and something like that was predominant. That said, I would recommend still streaming but at a coarser granularity.
Approving to let you revert quickly. If you want to patch-up instead of revert by limiting the buffering size, that is still an option.
|
I'd rather not destroy the connection on a badly formed header... I think I can rework the simplification to parse one element at a time, and that's probably preferable |
|
SGTM. That's my preference as well. |
|
Turning on auto-merge... I'm likely to be distracted later, and the sooner this is expunged from the build the better. |
…#26700)" (grpc#26825) This reverts commit 8bab3e4.
…#26700)" (grpc#26825) This reverts commit 8bab3e4.
2 participants
Reverts #26700
This PR introduces a dangerous overbuffering condition that's unsolvable with this approach.