On 5/4/16 7:47 PM, Sam Whited wrote:
On May 4, 2016 17:52, "Peter Saint-Andre" <[email protected]
<mailto:[email protected]>> wrote:
> The original suggestion (some years ago now) was browser-based
clients that could not easily download the entire Unicode character
database from a web server to the browser in order to do enforcement or
comparison,
That makes sense; the trie I use to store the derived properties for all
of Unicode 8.0.0 is 23 KB in memory (it would be a bit bigger over the
wire; though there's also room for improvement there). This seems
reasonable at first glance, however, a similar structure for
normalization ends up being 53 KB, and the one for width mapping is 11
KB; I suppose it begins to add up quickly if you're on a slow connection
That's an interesting approach. The person who originally brought this
up (I think it was Joe Hildebrand) might have been thinking that a web
client would need to import the entire UCD from a web server upon
initial connection, which would be prohibitive.
Peter
_______________________________________________
precis mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/precis
