src: initial localization support (work in progress)

[jasnell]

This is a work in progress effort to introduce the optional
ability to localize node's own outputs. Currently, Node (and V8)
outputs english only error, debug and help output. This is a step
towards allowing translated versions of node to be built.

Currently, this is hidden behind the --with-intl=full-icu switch.
This is because there are certain capabilities of ICU that are
only enabled with the full data set.

This is a work in progress that SHOULD NOT BE LANDED in master yet.

/cc @srl295 @mhdawson

[kkoopa]

The good thing about having only one language is that you can search the Internet for error messages and triage bug reports from users and such. It can already be difficult when errors come from the operating system With internationalization comes the need for some sort of identifier being printed for each message, so that one may know what that message actually is, but that also requires a lookup table for identifiers to their original messages. Does this functionality exist in ICU?

[jasnell]

ICU provides the message bundle mechanism. We would provide the actual lookup. It's not yet implemented within this PR, but my intent is to include the message identifier whenever a localized message is output. I was then planning to implement a simple userland utility installable from npm that would explain every error code with a localized description. e.g. something like...

$ npm install -g nexp
$ nexp NODE_ERR_001
...description printed here...

[kkoopa]

That sounds like a good plan. Just thought I'd mention this concern, but it is good that it has already been thought of.

On October 20, 2015 7:47:12 PM GMT+03:00, James M Snell notifications@github.com wrote:

ICU provides the message bundle mechanism. We would provide the actual
lookup. It's not yet implemented within this PR, but my intent is to
include the message identifier whenever a localized message is output.
I was then planning to implement a simple userland utility installable
from npm that would explain every error code with a localized
description. e.g. something like...

$ npm install -g nexp
$ nexp NODE_ERR_001
...description printed here...

---

Reply to this email directly or view it on GitHub:
#3413 (comment)

[rvagg]

Quick comment while I'm scanning the code -- maybe the l10n directory should be in the root rather than deps which are mostly things that are external and we don't normally touch unless we get upstream updates. Unless the intention is for l10n to exist in a separate repo and actually be a proper dep (might be a good idea!)?

[jasnell]

Yep, that will need to be worked out. One of the reasons I put it into deps is because of the optional ICU dependency. If ICU is not included in the build, the l10n dep is not built and linked in. However, because of the need for keeping the messages in sync with the source, having it out as a dependency doesn't make the most sense in the long run. As I said on the TC call, we need to figure out the "least annoying" way to set this up.

As an aside: right now ICU support is still optional in the build by default. It may be time to revisit that and make ICU build by default (e.g. make --with-intl=small-icu a default.