Re: [Help-librejs] Webpack plugin proposal to generate Web Labels page

[Antoine Lambert]

> Looks great!  I was also very pleased to found out, that it’s functional 
> without any client scripts (well, much more functional than Gitlab, at least 
> :-).

Indeed, the raw archive content can be browsed without JavaScript enabled but 
some UI features
will not be available (code highlighting, diff computation, ...). There is also 
some layout issues
as JavaScript is used to adjust some CSS dynamically. Nevertheless, this should 
be feasible
to propose a better navigation and design without JavaScript enabled (for 
browsing the
raw archived objects: directories, contents, revisions).

> Imho, that is quite feasible for the aims of propaganda, but for practical 
> freedom it would better have _working_ sourcemaps first.  They seem to be 
> there, but the declaration is nonstandard (just 
> ‘webapp.1328be1766de4e979da7.js.map’ instead of 
> ‘sourceMappingURL=webapp.1328be1766de4e979da7.js.map’), thus is not 
> recognized by Firefox.
>
> (By the way, it would be nice to have them for minified stylesheets too.)
Thanks for spotting this, there was some issues in my webpack configuration 
that was preventing correct loading
of js/css sourcemaps. This is now fixed and deployed.

> That’s by all means cool, but as a mere passerby, I am convinced, that 
> LibreJS have to make use of sourcemaps on its own, without mandating use of 
> another protocol.
Indeed, sourcemaps could be consumed by LibreJS to extract license information 
of the bundled source files.
Nevertheless, this requires that a license header is present in those. There 
exist cases where such headers
are not available. For instance in the core-js library [1], used in swh-web as 
polyfills provider (I try to target a wide
range of browsers in order for the application to run correctly everywhere), 
there is no license header in the
source files. So for those cases, the Web Labels approach seems the only way to 
state the licenses.

> I am not sure, if any of LibreJS developers actually read this (user’s) list. 
>  It might make sense to crosspost to <address@hidden>.
I do not consider my first message as a bug report so let's keep that thread 
alive here.

[1] https://github.com/zloirock/core-js

-- 
Antoine Lambert
Software engineer for Software Heritage https://www.softwareheritage.org
Inria, Paris
Github account: https://github.com/anlambert

signature.asc
Description: OpenPGP digital signature