<html>
<head>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">On 8/24/2015 1:35 PM, R Kent James
wrote:<br>
</div>
<blockquote cite="mid:55DB63E7.50801@caspia.com" type="cite">
<meta http-equiv="content-type" content="text/html; charset=utf-8">
We need to do some sort of announcement in the Thunderbird blog
about our plans concerning addons. I'd like to have feedback from
folks to see if there is any debate about what is the correct
direction for us.<br>
<br>
We've at least agreed that we are continuing to support binary
addons. Concerning signing, we took steps months ago to not move
forward on requiring addons to be signed, so there are no current
plans to require signing. There is still some debate about that in
bug 1168571. We should probably come to a firm decision and
announce it. Most commenters were opposed to signing, though there
were some holdouts.<br>
<br>
Then there is "<a moz-do-not-send="true"
href="https://blog.mozilla.org/addons/2015/08/21/the-future-of-developing-firefox-add-ons/">The
Future of Developing Firefox Add-ons - The Mozilla Blog</a>"
that announces the complete disabling of current XUL addons at
some point in the future. Several Thunderbird community members
commented on that blog post, strongly opposed to that direction.<br>
<br>
Contrast that with <a moz-do-not-send="true"
href="https://wiki.mozilla.org/Firefox/Go_Faster">Firefox/Go
Faster</a> where there are plans to expand the use of addons in
Firefox, adding so-called "system add-ons" and moving Hello to
one. (This is similar to what we are doing with Lightning, which
should hopefully make our Lightning integration easier in the
future).<br>
<br>
At this point, I think that the prevailing viewpoint is probably
the following, and I would like to announce this if possible in a
blog post:<br>
<br>
1) Thunderbird continues to support binary addons.<br>
2) Thunderbird will not require addon signing.<br>
3) Thunderbird has no current plans to disable the use of
traditional XUL/XPCOM addons in Thunderbird.<br>
<br>
This policy must be modified by the caveat "as long as core
Mozilla code can be used to support it".<br>
<br>
(I might also note that initial patches are being looked at for
the integration of the technology formerly know as Skinkglue into
Thunderbird core, to be called JsAccount, which makes it possible
to define new account types in Thunderbird using a traditional
XUL/XPCOM/JavaScript addon. This will almost certainly be in our
next major release).<br>
<br>
Could I have some comments or discussion on these proposed
positions?<br>
</blockquote>
<br>
On the topic of binary addons and XUL/XPCOM:<br>
<br>
To some degree, we are constrained by the infrastructure that the
core XPCOM or toolkit code will provide for us. If they completely
remove the functionality from the code, there is very little that we
can do, since we likely don't have the bandwidth to maintain code
ourselves. We are saved somewhat by the fact that, if B2G partner
repacks demand this functionality, the core people can't do much but
acquiesce, which I suspect gives us some breathing space, but it's
not exactly a permanent solution. So we do need to make it clear
that these technologies are effectively considered deprecated.<br>
<br>
In particular, for binary addons, we have three main binary addons
of any note (EWS, Lightning, and the libpurple chat). I think that
once none of those are using binary addons, we should seriously
consider flipping the switch, and if any other binary addons do
exist, the developers should make us aware of their existence.<br>
<br>
The question of XUL addons is harder to discuss, since I think that
Mozilla itself has no idea what to do about them. I do think it's
worth noting that where WebIDL exists for a feature, we should use
that instead of XPIDL interfaces; a good example would be charset
conversion.<br>
<pre class="moz-signature" cols="72">--
Joshua Cranmer
Thunderbird and DXR developer
Source code archæologist</pre>
</body>
</html>