netwerk/test/httpserver/TODO@ac0c01689b40 (annotated)
netwerk/test/httpserver/TODO
Thu, 15 Jan 2015 15:59:08 +0100
- author
- Michael Schloh von Bennewitz <michael@schloh.com>
- date
- Thu, 15 Jan 2015 15:59:08 +0100
- branch
- TOR_BUG_9701
- changeset 10
- ac0c01689b40
- permissions
- -rw-r--r--
Implement a real Private Browsing Mode condition by changing the API/ABI;
This solves Tor bug #9701, complying with disk avoidance documented in
https://www.torproject.org/projects/torbrowser/design/#disk-avoidance.
| michael@0 | 1 | Bugs to fix: |
| michael@0 | 2 | - make content-length generation not rely on .available() returning the entire |
| michael@0 | 3 | size of the body stream's contents -- some sort of wrapper (but how does that |
| michael@0 | 4 | work for the unscriptable method WriteSegments, which is good to support from |
| michael@0 | 5 | a performance standpoint?) |
| michael@0 | 6 | |
| michael@0 | 7 | Ideas for future improvements: |
| michael@0 | 8 | - add API to disable response buffering which, when called, causes errors when |
| michael@0 | 9 | you try to do anything other than write to the body stream (i.e., modify |
| michael@0 | 10 | headers or status line) once you've written anything to it -- useful when |
| michael@0 | 11 | storing the entire response in memory is unfeasible (e.g., you're testing |
| michael@0 | 12 | >4GB download characteristics) |
| michael@0 | 13 | - add an API which performs asynchronous response processing (instead of |
| michael@0 | 14 | nsIHttpRequestHandler.handle, which must construct the response before control |
| michael@0 | 15 | returns; |void asyncHandle(request, response)|) -- useful, and can it be done |
| michael@0 | 16 | in JS? |
| michael@0 | 17 | - other awesomeness? |
