HTTP2

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

HTTP2

Andrej van der Zee-4
Hi,

I understood that http2 is work in progress. Is there anything to say about when this might be released? 

Thank you,
Andrej

_______________________________________________
squid-users mailing list
[hidden email]
http://lists.squid-cache.org/listinfo/squid-users
Reply | Threaded
Open this post in threaded view
|

Re: HTTP2

Amos Jeffries
Administrator
On 28/02/19 6:30 am, Andrej van der Zee wrote:
> Hi,
>
> I understood that http2 is work in progress. Is there anything to say
> about when this might be released? 
>

No ETA sorry. Progress has slowed down so much I stopped even saying
"soon-ish" a year ago.

If anyone wants to jump in and lend a hand my HTTP/2 work is up on
github. IMO the best tasks to collaborate on would be designing cppunit
tests to ensure the code correctness of the HPACK coder, and/or to find
h2/h2c Parser bugs. Drop me a mail if that sounds interesting.

Amos
_______________________________________________
squid-users mailing list
[hidden email]
http://lists.squid-cache.org/listinfo/squid-users
Reply | Threaded
Open this post in threaded view
|

Re: HTTP2

Alex Rousskov
In reply to this post by Andrej van der Zee-4
On 2/27/19 10:30 AM, Andrej van der Zee wrote:

> I understood that http2 is work in progress.
> Is there anything to say about when this might be released? 

IMO, given the way the Squid Project operates right now, the correct
answer to that question is close to "hopefully not in the foreseeable
future": We cannot add quality HTTP/2 support right now, and adding some
hacky version of it would be disastrous for Squid stability, support,
and development. Combined with where the popular clients and origin
servers are going, it may be better to fantasize about HTTP/3 support
instead.

Based on Factory experience with adding HTTP/2 support to Web Polygraph,
I consider the following (partially overlapping) preconditions as
necessary for serious HTTP/2 (or HTTP/3) work in Squid:

  1. Proper QA infrastructure.

  2. Elimination of technical debt that prevents proper restructuring
     of HTTP/2-sensitive code.

  3. An agreement regarding overall HTTP/2 code architecture.

  4. An efficient way to accept huge code changes.

  5. A project lead capable, willing, trusted, and funded
     to orchestrate such a big change from beginning to end.

Right now, *none* of the above preconditions are satisfied.

There is slow but steady progress with #1 and areas of #2.

The situation with #3 and #4 is worse than it was a few years ago -- we
are wasting insane amounts of time on getting much simpler code changes
reviewed and accepted. Many changes require a rewrite before they should
be accepted (and some are indeed rewritten). Nobody can afford to
rewrite a pull request with initial HTTP/2 support!

We have nobody who can satisfy #5 criteria right now.


On 2/27/19 7:27 PM, Amos Jeffries wrote:

> If anyone wants to jump in and lend a hand my HTTP/2 work is up on
> github. IMO the best tasks to collaborate on would be designing
> cppunit tests
Creating more unofficial code is a bad idea at this time IMO.

Alex.
_______________________________________________
squid-users mailing list
[hidden email]
http://lists.squid-cache.org/listinfo/squid-users
Reply | Threaded
Open this post in threaded view
|

Re: HTTP2

Andrej van der Zee-4
Hi Alex / Amos,

It sounds like its still a long way to get HTTP/2 support released and contributing therefore is not an option in company time. 

Maybe a bit blunt, in our case it means looking for either Squid-workarounds for HTTP/2 servers we have to address (Apple APNS notification server for example), or change to another proxy with HTTP/2 support.

Best regards,
Andrej

 


On Thu, Feb 28, 2019 at 6:28 PM Alex Rousskov <[hidden email]> wrote:
On 2/27/19 10:30 AM, Andrej van der Zee wrote:

> I understood that http2 is work in progress.
> Is there anything to say about when this might be released? 

IMO, given the way the Squid Project operates right now, the correct
answer to that question is close to "hopefully not in the foreseeable
future": We cannot add quality HTTP/2 support right now, and adding some
hacky version of it would be disastrous for Squid stability, support,
and development. Combined with where the popular clients and origin
servers are going, it may be better to fantasize about HTTP/3 support
instead.

Based on Factory experience with adding HTTP/2 support to Web Polygraph,
I consider the following (partially overlapping) preconditions as
necessary for serious HTTP/2 (or HTTP/3) work in Squid:

  1. Proper QA infrastructure.

  2. Elimination of technical debt that prevents proper restructuring
     of HTTP/2-sensitive code.

  3. An agreement regarding overall HTTP/2 code architecture.

  4. An efficient way to accept huge code changes.

  5. A project lead capable, willing, trusted, and funded
     to orchestrate such a big change from beginning to end.

Right now, *none* of the above preconditions are satisfied.

There is slow but steady progress with #1 and areas of #2.

The situation with #3 and #4 is worse than it was a few years ago -- we
are wasting insane amounts of time on getting much simpler code changes
reviewed and accepted. Many changes require a rewrite before they should
be accepted (and some are indeed rewritten). Nobody can afford to
rewrite a pull request with initial HTTP/2 support!

We have nobody who can satisfy #5 criteria right now.


On 2/27/19 7:27 PM, Amos Jeffries wrote:

> If anyone wants to jump in and lend a hand my HTTP/2 work is up on
> github. IMO the best tasks to collaborate on would be designing
> cppunit tests
Creating more unofficial code is a bad idea at this time IMO.

Alex.
_______________________________________________
squid-users mailing list
[hidden email]
http://lists.squid-cache.org/listinfo/squid-users


--
Andrej van der Zee
Oranje-Vrijstaatkade 49
1093KS Amsterdam
+31-(0)6-8133-9388

_______________________________________________
squid-users mailing list
[hidden email]
http://lists.squid-cache.org/listinfo/squid-users
Reply | Threaded
Open this post in threaded view
|

Re: HTTP2

Alex Rousskov
On 2/28/19 3:35 PM, Andrej van der Zee wrote:

> It sounds like its still a long way to get HTTP/2 support released and
> contributing therefore is not an option in company time. 

With enough will and resources, we can promptly overcome the obstacles I
have outlined. They are not insurmountable! Whether investing in that is
better than the alternatives depends on your (company) circumstances. I
just wanted to warn that "more code" is kind of the opposite of what we
need right now as far as HTTP/2 support is concerned.

Alex.


> On Thu, Feb 28, 2019 at 6:28 PM Alex Rousskov wrote:
>
>     On 2/27/19 10:30 AM, Andrej van der Zee wrote:
>
>     > I understood that http2 is work in progress.
>     > Is there anything to say about when this might be released? 
>
>     IMO, given the way the Squid Project operates right now, the correct
>     answer to that question is close to "hopefully not in the foreseeable
>     future": We cannot add quality HTTP/2 support right now, and adding some
>     hacky version of it would be disastrous for Squid stability, support,
>     and development. Combined with where the popular clients and origin
>     servers are going, it may be better to fantasize about HTTP/3 support
>     instead.
>
>     Based on Factory experience with adding HTTP/2 support to Web Polygraph,
>     I consider the following (partially overlapping) preconditions as
>     necessary for serious HTTP/2 (or HTTP/3) work in Squid:
>
>       1. Proper QA infrastructure.
>
>       2. Elimination of technical debt that prevents proper restructuring
>          of HTTP/2-sensitive code.
>
>       3. An agreement regarding overall HTTP/2 code architecture.
>
>       4. An efficient way to accept huge code changes.
>
>       5. A project lead capable, willing, trusted, and funded
>          to orchestrate such a big change from beginning to end.
>
>     Right now, *none* of the above preconditions are satisfied.
>
>     There is slow but steady progress with #1 and areas of #2.
>
>     The situation with #3 and #4 is worse than it was a few years ago -- we
>     are wasting insane amounts of time on getting much simpler code changes
>     reviewed and accepted. Many changes require a rewrite before they should
>     be accepted (and some are indeed rewritten). Nobody can afford to
>     rewrite a pull request with initial HTTP/2 support!
>
>     We have nobody who can satisfy #5 criteria right now.
>
>
>     On 2/27/19 7:27 PM, Amos Jeffries wrote:
>
>     > If anyone wants to jump in and lend a hand my HTTP/2 work is up on
>     > github. IMO the best tasks to collaborate on would be designing
>     > cppunit tests
>     Creating more unofficial code is a bad idea at this time IMO.
>
>     Alex.
>     _______________________________________________
>     squid-users mailing list
>     [hidden email]
>     <mailto:[hidden email]>
>     http://lists.squid-cache.org/listinfo/squid-users
>
>
>
> --
> Andrej van der Zee
> Oranje-Vrijstaatkade 49
> 1093KS Amsterdam
> +31-(0)6-8133-9388
> https://www.linkedin.com/in/andrejvanderzee/

_______________________________________________
squid-users mailing list
[hidden email]
http://lists.squid-cache.org/listinfo/squid-users