7 ms·

Trekking the Road from SPDY to HTTP/2 in Firefox

31 points by patrickmcmanus 14y ago

CJefferson 14y ago
One thing this doesn't seem to answer... Why bother when we already have SPDY?
- sp332 14y ago
  SPDY patches some of the problems in the more common use cases for HTTP for today's applications. It doesn't really change much of the spec, and it doesn't even pretend to address potential future use cases. Here's a recent criticism by the lead architect of Varnish https://news.ycombinator.com/item?id=4253538 https://news.ycombinator.com/item?id=4253538
  - gsnedders 14y ago
    On the other hand, SPDY never tried to address those issues: the aim was to create a more efficient over-the-wire transport layer for HTTP/1.1 semantics.
    - sp332 14y ago
      I was answering this question https://news.ycombinator.com/item?id=4346116 https://news.ycombinator.com/item?id=4346116 Your point is exactly why we shouldn't settle for SPDY becoming HTTP 2.0.
- aidenn0 14y ago
  Or HTTP over SCTP
  - mbelshe 14y ago
    As soon as Microsoft embeds SCTP in every Windows device, we might be able to start discussing this...
    - aidenn0 14y ago
      SCTP can be implemented in a shared library and tunneled through UDP. That actually works better since otherwise NATs kill you
- Osiris 14y ago
  Isn't the idea to have a formal standard owned by a standards body? Currently SPDY has not been accepted by any standards body as a formal standard. That's probably why HTTP/2.0 is going to be based on SPDY, an attempt to codify an emerging standard.
- wmf 14y ago
  Renaming SPDY to HTTP/2.0 is on the table, but some people disagree with some of SPDY's design choices and are hoping to fix them.
- gioele 14y ago
  Because SPDY has well known flaws (let's call them design choices) for example in caching. There have been many discussions in HN about these topics. Also, because there is no such a thing as "SPDY" what you are referring to is "SPDY 2 as implemented by Google Chrome and almost fully documented by Google". SPDY is a work in progress. If you want a real deployment to be possible, you need to have at least a fully documented basic core that has been discussed openly, not just an implementation from a corporation subject to changes whenever internal pressures require it.
  - sp332 14y ago
    SPDY is an IETF draft, working toward a ratified standard. https://tools.ietf.org/html/draft-mbelshe-httpbis-spdy-00 https://tools.ietf.org/html/draft-mbelshe-httpbis-spdy-00
    - gioele 14y ago
      > SPDY is an IETF draft, working toward a ratified standard. https://tools.ietf.org/html/draft-mbelshe-httpbis-spdy-00 https://tools.ietf.org/html/draft-mbelshe-httpbis-spdy-00 Now it is, before it was not. And that is exactly my point in my reply: you cannot take whatever you call "SPDY" at the moment (implementations + wiki pages) and call it "HTTP 2". You need to document it, propose it, discuss it openly, accept criticism, fixes and, if needed, radical changes. This is what is happening now in the HTTPbis working group and is a good thing, but it is not going to be a rubber-stamping process for SPDY as it is now, I hope.
- mbelshe 14y ago
  Standardization is a critical step. It verifies that we use open IP which is available to everyone, provides a open forum for discussion, and archives the protocol in a way all vendors can use to resolve disputes about how we communicate.

CJefferson 14y ago

One thing this doesn't seem to answer... Why bother when we already have SPDY?

sp332 14y ago
SPDY patches some of the problems in the more common use cases for HTTP for today's applications. It doesn't really change much of the spec, and it doesn't even pretend to address potential future use cases. Here's a recent criticism by the lead architect of Varnish https://news.ycombinator.com/item?id=4253538 https://news.ycombinator.com/item?id=4253538
- gsnedders 14y ago
  On the other hand, SPDY never tried to address those issues: the aim was to create a more efficient over-the-wire transport layer for HTTP/1.1 semantics.
  - sp332 14y ago
    I was answering this question https://news.ycombinator.com/item?id=4346116 https://news.ycombinator.com/item?id=4346116 Your point is exactly why we shouldn't settle for SPDY becoming HTTP 2.0.
aidenn0 14y ago
Or HTTP over SCTP
- mbelshe 14y ago
  As soon as Microsoft embeds SCTP in every Windows device, we might be able to start discussing this...
  - aidenn0 14y ago
    SCTP can be implemented in a shared library and tunneled through UDP. That actually works better since otherwise NATs kill you
Osiris 14y ago
Isn't the idea to have a formal standard owned by a standards body? Currently SPDY has not been accepted by any standards body as a formal standard. That's probably why HTTP/2.0 is going to be based on SPDY, an attempt to codify an emerging standard.
wmf 14y ago
Renaming SPDY to HTTP/2.0 is on the table, but some people disagree with some of SPDY's design choices and are hoping to fix them.
gioele 14y ago
Because SPDY has well known flaws (let's call them design choices) for example in caching. There have been many discussions in HN about these topics. Also, because there is no such a thing as "SPDY" what you are referring to is "SPDY 2 as implemented by Google Chrome and almost fully documented by Google". SPDY is a work in progress. If you want a real deployment to be possible, you need to have at least a fully documented basic core that has been discussed openly, not just an implementation from a corporation subject to changes whenever internal pressures require it.
- sp332 14y ago
  SPDY is an IETF draft, working toward a ratified standard. https://tools.ietf.org/html/draft-mbelshe-httpbis-spdy-00 https://tools.ietf.org/html/draft-mbelshe-httpbis-spdy-00
  - gioele 14y ago
    > SPDY is an IETF draft, working toward a ratified standard. https://tools.ietf.org/html/draft-mbelshe-httpbis-spdy-00 https://tools.ietf.org/html/draft-mbelshe-httpbis-spdy-00 Now it is, before it was not. And that is exactly my point in my reply: you cannot take whatever you call "SPDY" at the moment (implementations + wiki pages) and call it "HTTP 2". You need to document it, propose it, discuss it openly, accept criticism, fixes and, if needed, radical changes. This is what is happening now in the HTTPbis working group and is a good thing, but it is not going to be a rubber-stamping process for SPDY as it is now, I hope.
mbelshe 14y ago
Standardization is a critical step. It verifies that we use open IP which is available to everyone, provides a open forum for discussion, and archives the protocol in a way all vendors can use to resolve disputes about how we communicate.