Bug 212012 – [GTK] RFE: replace libicu by glib

• Component: WebKitGTK
• Status: RESOLVED WONTFIX

[Tomasz Kłoczko]

Looks like WebKit is uisng libicu on locale, charsets and time zones operations. As at the moment all that functions are provided by glib I think that it would be good to consider swithing from libicu to glib as WebKit alredy is heavyly using glib. This would allow memory usage and probably will make Webkit code simpler.

[Дилян Палаузов]

As mentioned at https://gitlab.gnome.org/GNOME/evolution-data-server/-/issues/216 WebKit includes JavaScript interpreter.  This interpreter is supposed one day to return for input:

  new Intl.RelativeTimeFormat('de', {numeric:'auto'}).format(0, 'day')

output

  heute

How do you propose to implement Intl.RelativeTimeFormat in JavaScript without using ICU?  Does glib offer `Relative Time` conversions?

[Adrian Perez]

Some of the newest Web-facing APIs (and in particular the JS Intl ones)
which rely on implementation behaviour provided by ICU. While the specs
themselves may not explicitly mention ICU, they are often just exposing
functionality present in the library; this has been mentioned recently
in this webkit-dev mailing list thread:

  https://lists.webkit.org/pipermail/webkit-dev/2020-April/031162.html

Therefore, we would need ICU anyway for those. If we are going to need
the library anyway, there is no point in trying to avoid using it.

If you need to reduce memory (or disk) usage, you can look into building
ICU with some features and/or data tables removed. For example the support
for EBCDIC encoding is in general unneeded in the context of a Web engine.
If you are trying to build an embedded system where you know in advance
which languages will be supported, you can also remove the data tables
for languages you don't need to support (for example: if your target is
the US market, remove the CJK/Thai/etc. break iterator dictionaries and
rule sets). If you want to go down this route, please check the ICU site,
which has some information on this topic:

  http://userguide.icu-project.org/icufaq#TOC-How-can-I-reduce-the-size-of-the-ICU-data-library-

[Tomasz Kłoczko]

I must honestly say that I have no idea how low level details in glib and icu looks like.
In reality I'm more software packager which operates on scale whole Linux distribution with some descent developers skills than blood and flesh software developer :)
My observations and conclusions are formed from that kind above each software projects point of view.

I see only that glib provides some level of some necessary routines and whole gnome ecosystem is happy with what is provided by glib.
WebKit and e-d-s are kind of exceptions.

If it is still something missing in glib which forces to use icu I think this is something which should be discussed with glib developers which actually wrote that part.

Nevertheless wherever already is possible to replace icu API by glib one IMO it should be done without waiting on actual glib related answer because I see that Webkit already is heading to more and more of using glib :)

Will try to pass the subject to glib developers because they help may significantly  help finish transition and/or finish implementing still some missing bits in glib :)

[Tomasz Kłoczko]

Just FTR: I've asked for some help of glib developers.

Hopefully they will join with they forces :P
https://gitlab.gnome.org/GNOME/glib/-/issues/2112

[Adrian Perez]

*** Bug 211983 has been marked as a duplicate of this bug. ***