Talk:Website evaluation of ywamkb.net

Logo
but but but but is it not prettier this way????? ;-) Kevin


 * Of course it is! You did invent this test, didn't you? :-)) --PitPat 17:16, 7 September 2009 (CEST)
 * And along the way I learnt much about the guidelines I didn't know about!!! --Kev-The-Hasty 21:30, 8 September 2009 (CEST)
 * True. Step by step ... but that's why your evaluation is cool: it gives you hints of areas where you didn't thought about before. --PitPat 08:57, 9 September 2009 (CEST)

Referer
http://www.sitedossier.com/site/www.ywamkb.net found some more Referers for www.ywamkb.net (10, 5 blogs + 4 YWAM sites). --PitPat 14:10, 9 September 2009 (CEST)
 * This is what I thought you would discover the first time around. I have spent time getting others to put links on their sites to YWAMKB as I have links to theirs in the Websites list bit. --Kev-The-Hasty 15:32, 10 September 2009 (CEST)
 * But why google doesn't show them? Apparently, these websites are not indexed very often, maybe. It does matter (that google is not aware of it), because google is rating the results mainly based on the number (and quality) of links to this site.--PitPat 16:41, 10 September 2009 (CEST)

Mind the gap
I can implement some proxy-caching which will reduce the php processing time but need assurance of benefit of effort invested. Input please!!!! --Kevin


 * I don't think this is the bottleneck here. An APC-Cache, memcache, reverse-proxy etc. would be reasonable if the server is under such a big hit (i.e. regular CPU peeks at 100%, or similar) because of many people accessing it. Here the bottleneck is the 56k-Modem - and the reason why the site is calculating 30 seconds is that these connections tend to have a very high latency (ping time). Which means, that for each image the browser requests, this latency is added 2 times: the browser waits until the previous image is loaded completely, than starts to load the next image. On the client side, you could change some preferences (higher MTU, more connections per site, make sure the browser speaks HTTP/1.1), but on the server side, you can't do very much unless:
 * Ensure the server speaks HTTP/1.1 (apache does)
 * Send content gzipped (see mod_deflate. This server does for some files. (Do the websiteoptisation Test to see this.)
 * Reduce the number of images. Esp. main page has many images. CSS-Images also count in this report, even if not used.
 * Combine the (7!) CSS files to one big one, maybe compress it. This can happen offline (e.g. by deployment script.) If some lines of CSS are included by plugins, prefer to write it into the HTML?
 * Each img-Tag has to have a height / width tag. (Not always the case)
 * Serve a compressed version of HTML (look at the HTML code of google to see what I mean.)


 * You can test a pseudo modem connection with iprelay. Another good test for client caching is Firebug, a Firefox extension.
 * --PitPat 14:55, 12 September 2009 (CEST)


 * This is excellent. I just used it and it is as I suspected. I throttled bandwidth to 7000B/s (=56Kps I believe). Y-slow extension (Firefox) told me Main page is 23 seconds. Mostly Javascript I guess as it pauses partially displaying the page before completing. BUT, next page is about 1.5 s and so on. A page with a picture might take 6 more seconds to download. Thus I think the browser does a good job of caching the javascript and page furniture. This is re-assuring for me that the Mind The Gap value is preserved. Perhaps improvements can be made with slimming the javascript or getting mod_deflate to work properly! However, it is the 2nd and 3rd page where real speed savings are made. I am satisfied. Comments? --Kev-The-Hasty 14:46, 18 September 2009 (UTC)
 * Sorry to disappoint you. Remembering my dial-up times, I rarely had more than 4,0 kB/sec (that is, 4096 B/s) down. I think, 56kbit (= 5,6 kB )is the theoretical limit, as 100Mbit/s at Ethernet, which will never be reached. It got a bit better after tweaking the windows registry (MTU and similar). --PitPat 09:40, 19 September 2009 (UTC)
 * How quickly we forget! I think 4500 to 5000 B/s was normal for where I was. We once had a 28Kps modem that we shared from our Lan with 4 people on it. Strangly it was never criticised for slow bandwidth but often for the lag whilst it dialed up. --Kev-The-Hasty 16:55, 19 September 2009 (UTC)
 * I will run some more figures and perhaps run some of the static javascript through a minimiser to shed some weight. I think the test metric should be 4000B/s. (just to be conservative) and the average of 5 pages? (top 5 headings maybe?). We could also use this to create a league table of YWAM sites! --Kev-The-Hasty 16:55, 19 September 2009 (UTC) Mind The Gap Test.