Name: Compact Language Detection 2
Short Name: cld_2
URL: https://code.google.com/p/cld2/
Version: 0
License: Apache 2.0
Security Critical: yes
Description:
The CLD is used to determine the language of text. In Chromium, this is used
to determine if Chrome should offer Translate UX to the user.
Dynamic Mode
============
Prior to CLD2's trunk@155, Chromium has always built CLD2 statically. The data
needed for CLD2 to perform its language detection has been compiled straight
into the binary. This contributes around 1.5 megabytes to the size of Chrome
and embeds one or more large rodata sections to the executable.
Starting with CLD2's trunk@r155, there is a new option available: dynamic mode.
In dynamic mode, CLD2 is built without its data; only the code is compiled, and
the data must be supplied at runtime via a file or a pointer to a (presumably
mmap'ed) read-only region of memory.
Tradeoffs to consider before enabling dynamic mode:
Pros:
* Reduces the size of the Chromium binary by a bit over a megabyte.
* As the data file rarely changes, it can be updated independently.
* Depending upon the update process on your platform, this may also reduce
the size of Chromium updates.
* It is possible to run Chromium without CLD2 data at all (language
detection will always fail, but fails gracefully).
* Different types of CLD2 data files (larger and more accurate or smaller
and less accurate) can be dynamically downloaded or chosen depending
on runtime choices.
Cons:
* Data files must be generated and checked into source control by hand.
* At runtime a data file must be opened and its headers parsed before CLD2
can be used in any given process (this time should be negligible in most
circumstances). This will prevent language detection from working until
a data file has been loaded.
To enable dynamic mode, set 'cld_dynamic' to '1' in ../../build/common.gypi.
For more information on the changes required to use dynamic mode, see:
https://codereview.chromium.org/187393005