[Cosmo-dev] CosmoUI Timezone Proposal

[Matthew Eernisse]

Comment below.

Bobby Rullo wrote:
> That's wrong actually. Parsing takes quite a bit of time. Your original 
> code split files into lines, but didn't do any parsing until that 
> timezone was called for. My code which does the actual parsing during 
> load is actually pretty slow - takes like 5 seconds for the actual parse.

What is the point of changing the implementation so that it's now way 
slower than it was? The initial implementation loaded pretty much 
instantaneously, and as far as I could see it took no perceptible time 
at all to calculate offsets for TZ/dates on the fly. Do we get some 
other practical benefits from doing that?

> All the timezones combined and stripped of comments would be about 50k - 
> 100k tops,  way less gzipped. And they get cached. So size really is not 
> a problem here.

Size is less of a problem for us, since we are doing both the library 
and the app code. We can kind of do whatever we want. But for the Dojo 
guys, size/download-time is at an *incredible* premium.

And 'cache' doesn't necessarily help that much with the file-size issue, 
since browsers in general (IE of course has it's own screwy caching 
rules) flush cache when you first fire them up, so you still have to 
pull the stuff down every time you log into the app. There are 
server-side things you can do to attempt to mitigate some of that, but 
Dojo is supposed to be an all-client-side solution.


Matthew