Feedback from first read of QUIC/HTTP draft 08

[Mikkel Fahnøe Jørgensen <[e-mail]>]

This was originally #1001, but moved here on upon request.

Reading through the QUIC/HTTP-draft-08 a few things appeared unclear. This
is just a summary based on my first pass over the document, so take it for
what it is. I have only superficial knowledge of HTTP/2 in advance.

   1. A bit too much discussion of HTTP/2 instead of leaving that for
   section 8. It does not help understanding QUIC/HTTP from scratch. And
   perhaps a bit too much assumption about HTTP/2 knowledge while implementers
   might have skipped that step in their evolution.
   2. No explanation of how methods are communicated. Header fields
   reference HTTP 1.1 draft, but from skimming HTTP/2 doc, it is clear that
   these (as suspected) are transfered via special headers known as pseudo
   headers).
   3. Lots of frame type flags, none used, and only a few types. Why not
   either remove flags and have more types, or have a type bit encoding for
   flags presence (this may be due to early state of draft, just mentioning).
   4. Why are settings 16 bits instead of varints, and why is the space so
   large. A 256 or even 2K space would be easy to have in a table, 64K
   requires a map or switching logic. A settings space still requires that
   both ends understands a signficant portion of settings to make any real
   sense. Who decides to add more settings. Is this to have space for the next
   century, or just until the next QUIC version?
   5. There is an End Header Block something, but it is unclear what it is.
   Is it a flag to be defined, or some pseudo header?
   6. The GOWAY discussion could probably be shortened and made it bit
   clearer - it takes a few readings to understand the point of multiple GOWAY
   frames and why the ID can decrease (not sure I do fully understand).
   7. A basic example of a short HTTP 1.1 request and how it looks in the
   protocol would be helpful.
   8. The HPACK, to be replaced, doc is hard to parse, and it doesn't sound
   like it is as complicated as it reads. Hopefully a future QUIC equivalent
   would be a bit easier to digest.
   9. The entire push hierarchy sounds very complex - I wonder what the
   takeaway from HTTP/2 is - is it really worthwhile, or is no-one using it
   yet, but in 10 years it will be huge?