%HTMLlat1; %HTMLsymbol; %HTMLspecial; ]> Towel-snapping optimisation
Roland van Ipen­burg
To be stolen or blogged

Tow­el-snap­ping op­ti­mi­sa­tion

Mon­day 13 Fe­bru­ary 2012 22:44

When your au­di­ence is most­ly n00b front-end strug­glers, as a medi­um ex­ist­ing be­cause of pub­li­ca­tions and not so­lu­tions it's per­fect­ly rea­son­able to waste every­one's time adding an­oth­er chap­ter to the cir­cle jerk that do­ing some Bos­ton Globe web­site ap­par­ent­ly was. Other­wise it makes no sense, for the fol­low­ing rea­sons:

There is no re­li­able way to de­ter­mine the amount of band­width a user is will­ing to spend on a sin­gle re­source. Be­cause screen size is all we've got to query in some mo­bile cu­ri­ous way that doesn't mean it's us­able or ef­fi­cient to do so. So it's some Spiel­erei to pre­pare for the mag­ic event when there is an API that we can use to de­ter­mine the band­width on a de­vice? No, be­cause mo­bile means there isn't a sin­gle band­width on a de­vice. Even an API can't pre­dict when con­di­tions sud­den­ly get worse be­cause a de­vice moves from Wifi to Edge. In such a volatile en­vi­ron­ment it's back­ward to keep think­ing in mono­lith­ic bitmapped re­sources ar­riv­ing af­ter craft­ing a sin­gle re­quest for it. There's wavelets, pro­gres­sive schemes and vec­tor based im­ages that are much bet­ter at us­ing band­width flex­i­ble and ef­fi­cient­ly be­cause they can stream un­til some­thing de­ter­mines there is enough res­o­lu­tion, and re­sume when more is need­ed. But if you're look­ing for a markup so­lu­tion - be­cause that is what your au­di­ence is crav­ing for - it's just mov­ing the same old pro­to­col into an­oth­er lay­er, ef­fec­tive­ly of­fer­ing the user no ad­van­tage. Just wast­ing time wait­ing to get some­thing into a stan­dard it shouldn't be in in the first place, but ap­par­ent­ly it's to hard to get de­sign­er-slash-de­vel­op­ers to ask serv­er spe­cial­ists to solve their prob­lems prop­er­ly. That's why hey are called HTML mon­keys.

And even if you've gone through such great lengths to pre­vent suck­ing re­dun­dant data over some­one's pre­cious con­nec­tion, only one click seper­ates that user from go­ing to an­oth­er web­site that didn't and makes the ef­fect of your op­ti­mi­sa­tions seem pret­ty use­less. It's like risk­ing your life to re­move a sin­gle mine from a mine­field and then claim you've made an im­prove­ment in safe­ty for the user. But that user isn't go­ing near your sin­gle safe spot as long as all the oth­er mines are still there. Nice tech demo, but in the real world the whole ecosys­tem the user op­er­ates in has to have some min­i­mal lev­el to make it us­able in he first place. And the core of that ecosys­tem is still the brows­er.

Browsers should do a lot more than just im­ple­ment web stan­dards flaw­less­ly. Ini­tia­tives like Opera Mini show users can be giv­en good tools with­out some stan­dards body first have to make a stan­dard for it, or every web­mas­ter come up with his own rein­ven­tion of what should be brows­er fun­cional­i­ty. But if all HTML mon­keys can do is markup and think in markup they will nev­er un­der­stand how to build a brows­er and solve the real prob­lems. They seem to have enough is­sues grasp­ing why browsers don't func­tion as they would ex­pect them to work from their markup view­point any­way.


Book­mark this on De­li­cious

Add to Stum­bleUpon

Add to Mixx!



application browser cool data days different flash game gta html ibook internet linux movie open play playstation run screen server side single site stuff system train user web windows work wrong

Blog Posts (421)

Image Gal­leries

ipen­bug Last.fm pro­file

ipen­bug last.fm pro­file

Fol­low me on Twit­ter

Roland van Ipen­burg on face­book
Lin­ux Regis­tered User #488795
rolipe BOINC com­bined stats


Add to Google

Valid XHTML + RFDa Valid CSS! Hy­phen­at­ed XSL Pow­ered Valid RSS This site was cre­at­ed with Vim Pow­ered by Bri­co­lage! Pow­ered by Post­greSQL! Pow­ered by Apache! Pow­ered by mod­_perl! Pow­ered by Ma­son! Pow­ered by Perl Made on a Mac Pow­ered By Mac OS X XS4ALL This site has been proofed for ac­cu­ra­cy on the VISTAWEB-3000 Creative Com­mons Li­cense
This work by Roland van Ipen­burg is li­censed un­der a Creative Com­mons At­tri­bu­tion-Non­com­mer­cial-Share Alike 3.0 Un­port­ed Li­cense.
Per­mis­sions be­yond the scope of this li­cense may be avail­able at mail­to:ipen­burg@xs4all.nl.