librelist archives

« back to archive

Opinions about SPDY

Opinions about SPDY

From:
Mathijs Kwik
Date:
2011-08-17 @ 15:07
Hi all,

I haven't looked at SPDY much, other than reading some descriptions of
it, and checking which google sites are already using it (almost all).
It all sounds very exciting and promising, and it seems like a perfect
carrier for async/multiplexed traffic.
Also, I got spammed today about a full book just about SPDY (
http://pragprog.com/book/csspdy/the-spdy-book ).

But I was excited about websockets as well, until I read Zed's opinion.
So have people here actually used it already? Looked at RFCs?

Is it all glory and without stupid/problematic parts?
Can we expect a mongrel2 frontend someday?

Thanks,
Mathijs

Re: [mongrel2] Opinions about SPDY

From:
Zed A. Shaw
Date:
2011-08-17 @ 16:12
On Wed, Aug 17, 2011 at 05:07:40PM +0200, Mathijs Kwik wrote:
> Hi all,
> Is it all glory and without stupid/problematic parts?
> Can we expect a mongrel2 frontend someday?
> 

Remember when Microsoft made IE work better with IIS by skipping the
initial TCP setup?  Remember how everyone went apeshit that they would
abuse their monopoly power to unilaterally make changes to
specification?  Remember how everyone screamed this was "embrace and
extend"?

Yeah, same thing Google's doing, and I won't support it.

If they manage to:

1. Get it in officially in Firefox and IE by getting it approved by the
W3C.
2. Write a reference implementation that is not a pile of nasty C++ code
pretending to be C.
3. Get Apache to adopt it.
4. Have more metrics than just their "it saved us 3%" to justify it.
5. Sign a covenant saying that they will pay a charity $1million for
every day that Chrome deviates from the released protocol and reference
implementations.

Then I'll probably throw it in, but by then hell will have frozen over
and I probably wouldn't need to work on Mongrel2 anymore.

-- 
Zed A. Shaw
http://zedshaw.com/

Re: [mongrel2] Opinions about SPDY

From:
Mathijs Kwik
Date:
2011-08-17 @ 17:18
Hehe,

That's what opinion's should look like ;)
I agree on the embrace&extend part and needing more support (just firefox
will do for me though). Indeed some standards body should manage and guard
the specs, and a clear, nicely written/commented reference seems like a must
too.

So I agree for now, but I'm still somewhat hopeful that this becomes more
standardized over timr, with more browsers backing it.

The thing I like is the ability to send more than was asked for. So a client
requesting an html page, can be given html, css, js, and images, without
additional roundtrips. This almost removes the need for asset packaging,
sprites, and other optimizations.

anyway, thanks for your vision on this matter.

Mathijs
Op 17 aug. 2011 18:13 schreef "Zed A. Shaw" <zedshaw@zedshaw.com> het
volgende:
> On Wed, Aug 17, 2011 at 05:07:40PM +0200, Mathijs Kwik wrote:
>> Hi all,
>> Is it all glory and without stupid/problematic parts?
>> Can we expect a mongrel2 frontend someday?
>>
>
> Remember when Microsoft made IE work better with IIS by skipping the
> initial TCP setup? Remember how everyone went apeshit that they would
> abuse their monopoly power to unilaterally make changes to
> specification? Remember how everyone screamed this was "embrace and
> extend"?
>
> Yeah, same thing Google's doing, and I won't support it.
>
> If they manage to:
>
> 1. Get it in officially in Firefox and IE by getting it approved by the
> W3C.
> 2. Write a reference implementation that is not a pile of nasty C++ code
> pretending to be C.
> 3. Get Apache to adopt it.
> 4. Have more metrics than just their "it saved us 3%" to justify it.
> 5. Sign a covenant saying that they will pay a charity $1million for
> every day that Chrome deviates from the released protocol and reference
> implementations.
>
> Then I'll probably throw it in, but by then hell will have frozen over
> and I probably wouldn't need to work on Mongrel2 anymore.
>
> --
> Zed A. Shaw
> http://zedshaw.com/

Re: [mongrel2] Opinions about SPDY

From:
David Hofmann
Date:
2012-04-15 @ 16:29
Hello,

Just an update on the thread

AFAIK spdy was submitted to IETF
Google and Twitter are officially using it
Chrome and Mozilla Firefox browsers supports it
Jetty web server supports it
Apache httpd has mod_spdy
Nginx support is comming

Mongrel2 ... ?



On Wed, Aug 17, 2011 at 1:18 PM, Mathijs Kwik <bluescreen303@gmail.com>wrote:

> Hehe,
>
> That's what opinion's should look like ;)
> I agree on the embrace&extend part and needing more support (just firefox
> will do for me though). Indeed some standards body should manage and guard
> the specs, and a clear, nicely written/commented reference seems like a
> must too.
>
> So I agree for now, but I'm still somewhat hopeful that this becomes more
> standardized over timr, with more browsers backing it.
>
> The thing I like is the ability to send more than was asked for. So a
> client requesting an html page, can be given html, css, js, and images,
> without additional roundtrips. This almost removes the need for asset
> packaging, sprites, and other optimizations.
>
> anyway, thanks for your vision on this matter.
>
> Mathijs
> Op 17 aug. 2011 18:13 schreef "Zed A. Shaw" <zedshaw@zedshaw.com> het
> volgende:
>
> > On Wed, Aug 17, 2011 at 05:07:40PM +0200, Mathijs Kwik wrote:
> >> Hi all,
> >> Is it all glory and without stupid/problematic parts?
> >> Can we expect a mongrel2 frontend someday?
> >>
> >
> > Remember when Microsoft made IE work better with IIS by skipping the
> > initial TCP setup? Remember how everyone went apeshit that they would
> > abuse their monopoly power to unilaterally make changes to
> > specification? Remember how everyone screamed this was "embrace and
> > extend"?
> >
> > Yeah, same thing Google's doing, and I won't support it.
> >
> > If they manage to:
> >
> > 1. Get it in officially in Firefox and IE by getting it approved by the
> > W3C.
> > 2. Write a reference implementation that is not a pile of nasty C++ code
> > pretending to be C.
> > 3. Get Apache to adopt it.
> > 4. Have more metrics than just their "it saved us 3%" to justify it.
> > 5. Sign a covenant saying that they will pay a charity $1million for
> > every day that Chrome deviates from the released protocol and reference
> > implementations.
> >
> > Then I'll probably throw it in, but by then hell will have frozen over
> > and I probably wouldn't need to work on Mongrel2 anymore.
> >
> > --
> > Zed A. Shaw
> > http://zedshaw.com/
>

Re: [mongrel2] Opinions about SPDY

From:
Nathan Duran
Date:
2012-04-16 @ 05:23
On Apr 15, 2012, at 9:29 AM, David Hofmann wrote:

> Hello,
> 
> Just an update on the thread
> 
> AFAIK spdy was submitted to IETF
> Google and Twitter are officially using it
> Chrome and Mozilla Firefox browsers supports it
> Jetty web server supports it
> Apache httpd has mod_spdy
> Nginx support is comming
> 
> Mongrel2 ... ?

Until it's in Safari and IE it's not really worth the effort. Has anyone 
proven it actually works well enough to bother with? 

Re: [mongrel2] Opinions about SPDY

From:
anuj dutta
Date:
2012-04-16 @ 06:54



On 16 Apr 2012, at 06:23, Nathan Duran <principal@khiltd.com> wrote:

> 
> On Apr 15, 2012, at 9:29 AM, David Hofmann wrote:
> 
>> Hello,
>> 
>> Just an update on the thread
>> 
>> AFAIK spdy was submitted to IETF
>> Google and Twitter are officially using it
>> Chrome and Mozilla Firefox browsers supports it
>> Jetty web server supports it
>> Apache httpd has mod_spdy
>> Nginx support is comming
>> 
>> Mongrel2 ... ?
> 
> Until it's in Safari and IE it's not really worth the effort. Has anyone
proven it actually works well enough to bother with? 

I think it should be implemented regardless of what browsers support it. 
Also, any clients not on spdy should just fallback to http.

Anuj

Re: [mongrel2] Opinions about SPDY

From:
Nathan Duran
Date:
2012-04-16 @ 06:11
On Apr 15, 2012, at 11:54 PM, anuj dutta wrote:

> I think it should be implemented regardless of what browsers support it.
Also, any clients not on spdy should just fallback to http.

So go implement it?

Re: [mongrel2] Opinions about SPDY

From:
Loic d'Anterroches
Date:
2012-04-16 @ 06:24
Hello,

On 2012-04-16 07:23, Nathan Duran wrote:
> On Apr 15, 2012, at 9:29 AM, David Hofmann wrote:
>> Hello,
>>
>> Just an update on the thread
>>
>> AFAIK spdy was submitted to IETF
>> Google and Twitter are officially using it
>> Chrome and Mozilla Firefox browsers supports it
>> Jetty web server supports it
>> Apache httpd has mod_spdy
>> Nginx support is comming
>>
>> Mongrel2 ... ?
> 
> Until it's in Safari and IE it's not really worth the effort. Has anyone
proven it actually works well enough to bother with? 

The real question is why would you need it? Adding a feature to a system
must answer a real need.

Google optimized its backend, front-end etc. to death before considering
that compressing the headers in a request could save a bit more time
with multiplexing helping for large web apps where they have assets all
over the place to download.

Nowadays, you can get the same for less. You package all your .js in one
minimized file, idem for CSS, you sprite your icons and carefully
managed expiration headers with the path to an asset containing the hash
of the content. You end up having something being able to answer with an
extremely low latency.

We have the numbers from Google on how "good" it is, I am really waiting
for numbers from other players using nginx to be able to assess if it is
really worth it. Only numbers will tell us.

loïc


Re: [mongrel2] Opinions about SPDY

From:
Ian Barber
Date:
2012-04-16 @ 07:07
On Mon, Apr 16, 2012 at 6:23 AM, Nathan Duran <principal@khiltd.com> wrote:

>
> Until it's in Safari and IE it's not really worth the effort. Has anyone
> proven it actually works well enough to bother with?
>

FWIW, Microsoft submitted their own HTTPbis protocol the other day as well
- it's basically SPDY without a couple of features. I would imagine that it
or something like it will start making it into IE eventually.

Ian

Re: [mongrel2] Opinions about SPDY

From:
Javier Guerra Giraldez
Date:
2012-04-16 @ 14:30
On Mon, Apr 16, 2012 at 2:07 AM, Ian Barber <ian.barber@gmail.com> wrote:
> I would imagine that it or something like it will start making it into IE
> eventually.

.... and it will only be worthy of consideration if it's _fully_
interoperable with everybody else's optimizations.

right now, there's a "proprietary but well documented" protocol.  if
tomorrow there are two that present even the slightliest trouble, it
will be better to stick with HTTP.

OTOH, if both varieties _do_ work together, then adopting it gets more
and more interesting.

-- 
Javier