[illumos-Developer] webrev: multibyte locale collation and license cleanups

[Kyle McDonald]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
 


On 9/17/2010 11:01 PM, Garrett D'Amore wrote:
> This is a big review, but its *really* only 2700 lines when you factor
> out the two files that are each over 20,000 lines. (The UnicodeData.txt
> file and the generated map file from it.)
>
> The review is here:
>
> http://mexico.purplecow.org/gdamore/webrev/collate/
>
>
> There are three bugs:
>
> c) The locale data for ctype and collate are binary data files. It is
> wasteful to use network byte ordering for them, since they are not
> shared across platforms. They are now native byte ordering. This
> should shave quite a few cycles for x86 boxes when loading these data
> files. Also, in the future, we can now consider using mmap() of these
> files (that is the subject of a future RFE) to get even more
> performance.
>
Are these files installed in /usr/share (or /usr/lib/share?) on the
user's system?

If yes, then they should be moved if they are no longer shareable
across platforms. It's not super common, but I do know of places that
still mount /usr/share on client machines from a single server. It
sounds like this breaks that.

  -Kyle

>
> Its a big review, and I'm grateful for any feedback I can receive.
>
> Thanks!
>
> - Garrett
>
>
> _______________________________________________
> Developer mailing list
> Developer at lists.illumos.org
> http://lists.illumos.org/m/listinfo/developer
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.14 (MingW32)
 
iQEcBAEBAgAGBQJMlMW7AAoJEEADRM+bKN5wRmYH/1KwUcKq0q2oRO6rz0ehYPn6
2KJpdZgyCsrwFalkrQvzyTfrpKPjWGqiqvBi4tfQ6NP9rB06gBFF50MTeiRyPa7y
exXIuXWicpMSTl7DC6yTO72ZC9/yDbVvBR8BvNAhXc6CgRMWie8kjUvwxcMCrWpv
OX/22P9XqyUvn1MtWDTfOW1ZPNAynW11D+1IsusGS6+d9nTcfBa/NK4gy0MPAcaK
+Fcc2Bm8iFsfuPz0kct7vm7NfVat/J4eZGVZc316NDR3OABgXnXaVKK4Ba49xRM8
GKDDgnwX6b47m0XCZBTJACuhhH//KwwvBGeUsA7wh9cCenptBY1LoyHe1N2Oi4c=
=NT9U
-----END PGP SIGNATURE-----