Path: csiph.com!usenet.pasdenom.info!dedibox.gegeweb.org!gegeweb.eu!nntpfeed.proxad.net!proxad.net!feeder2-2.proxad.net!newsfeed.arcor.de!newsspool3.arcor-online.net!news.arcor.de.POSTED!not-for-mail Content-Type: text/plain; charset="UTF-8" Message-ID: <2124180.lVgQKrcW98@PointedEars.de> From: Thomas 'PointedEars' Lahn Reply-To: Thomas 'PointedEars' Lahn Organization: PointedEars Software (PES) Date: Sun, 27 Oct 2013 20:16:22 +0100 User-Agent: KNode/4.10.5 Content-Transfer-Encoding: 8Bit X-Face: %i>XG-yXR'\"2P/C_aO%~;2o~?g0pPKmbOw^=NT`tprDEf++D.m7"}HW6.#=U:?2GGctkL,f89@H46O$ASoW&?s}.k+&. <3382075.QeW2rGUY4L@PointedEars.de> <69167318.yBAuBtPbmc@PointedEars.de> <4401911.7nMNhKR3Lp@PointedEars.de> MIME-Version: 1.0 Lines: 96 NNTP-Posting-Date: 27 Oct 2013 20:16:24 CET NNTP-Posting-Host: 5a4b07aa.newsspool1.arcor-online.net X-Trace: DXC=VfXI1E<7ol785[]]\]T081ic==]BZ:af>4Fo<]lROoR1nkgeX?EC@@0E^]7Dg`Z:Q:DZm8W4\YJNkjAfSmEgga? X-Complaints-To: usenet-abuse@arcor.de Xref: csiph.com comp.lang.javascript:21757 Eric Bednarz wrote: > Thomas 'PointedEars' Lahn writes: >>> http://www.browserscope.org/?category=network&v=top > ^^^^^^^ >> The test results are flawed. > > Of course they are. Why “of course”? >> Unsurpringly, that site is _not_ authored by Steve Souders > > | User: stevesou...@gmail.com > | Role: Owner > | Duties: > | * Lead by providing a project vision and roadmap > | * Have fun hacking and learn new stuff! > | * Component owner for "Network" tests > | Notes: Steve Souders > > https://code.google.com/p/browserscope/people/detail?u=114676337053000557002 I stand corrected; but see below. >> who consciously speaks about *servers* > > Ohhhhhhhhhhhhhm now I see. It's about *servers*, is it. So all you need > is a single hostname (or domain name, if you prefer) and a load balancer > to offload every request to a different *server*. And there you have it, > unlimited parallel connections. You are not making sense. Most certainly the number of HTTP connections an HTTP client can make is limited. >> (or hosts, if you >> prefer), _not_ host *names*. > > Let's take a look in his latest blog post. > > | You don’t need to setup different servers for each domain – just > | create CNAMEs. The browser doesn’t care about the final IP address – > | it only cares that the hostnames are different. > > http://www.stevesouders.com/blog/2013/09/05/domain-sharding-revisited/ No rationale is being given there as to why HTTP connections should have anything to do with host names. A few years earlier… pp. | It’s likely that Firefox 3 will support 6 connections per server in an | upcoming beta release, which means more discussion is expected. | Section 8.1.4 of the HTTP/1.1 RFC says a “single-user client SHOULD NOT | maintain more than 2 connections with any server or proxy. | Opera and Safari hold that honor supporting 4 connections per server. | The table below shows the number of connections per server supported by | current browsers for HTTP/1.1 as well as HTTP/1.0. | | Browser HTTP/1.1 HTTP/1.0 | […] | Opera 10.51+ 8 ? Compare that to the current table at BrowserScope: | name score PerfTiming Connections per Hostname | Max Connections | […] | Opera 10 7/16 no 8 | 30 It is either the maximum number of connections per server/host or the maximum number of connections per hostname. It cannot be both. It is interesting, though, that there are two limits even per BrowserScope: the maximum number of connections “per Hostname” (or per _server_?) and the maximum number of (total) connections. Whatever the numbers, we must assume that an HTTP connection is a scarce resource. A developer should not use their users' resources needlessly. Wasting this resource in particular decreases the user experience because at some point other requests have to wait. Therefore, in general it cannot be recommended to use many static script resources. Which is why modular libraries usually come with a mechanism to avoid that, as Stefan has pointed out already. -- PointedEars Twitter: @PointedEars2 Please do not Cc: me. / Bitte keine Kopien per E-Mail.