Bug#678445: ITP: dunst -- minimalistic notification daemon

[Sascha Kruse <knopwob@googlemail.com>]

On Fri, Jun 22, 2012 at 01:26:59AM +0300, Timo Juhani Lindfors wrote:

Hi,

I'm the author of dunst.

> Michael Stapelberg <michael+db20090501@stapelberg.de> writes:
> >  Dunst is a minimalistic notification daemon for Linux. It's designed to
> >  fit nicely into minimalistic windowmanagers like dwm, but it should
> >  work on any Linux desktop.
>
> I think you should elaborate on what "minimal" means. Does it try to
> implement the specification or just make most common applications work?

The term minimal is used in regard to it's visual appearance. It's only
a colored box with (unformatted) text. I try to implement the whole
specification unless it conflicts with this. That being said, if I
understand the specifications correctly and images are optional it
should be possible to implement the whole (non-optional part) of the specification.

Currently the whole specification isn't met, though. Most noticeably dunst
ignores:

    - the replaces_id (Table 6. Notify Parameters)

    - 9.1.3 CloseNotification  *

    - 9.2 Signals

    *  Dunst sends a reply to the client (otherwise some clients become
       unresponsive (until some timeout runs out!?) but it doesn't
       actually close the notification.

So the current state is "make most common applications work" and the
things listed above can be considered as bugs.



> I looked at the code and it seems to support only the "body" capability.
>
> It does not support "body-markup" but still contains stuff like
>
>     str = string_replace("&quot;", "\"", str);
>     str = string_replace("&apos;", "'", str);
[ ..snip..]
>     str = string_replace("</u>", "", str);
>     str = string_replace("</a>", "", str);
>
>     start = strstr(str, "<a href");
>
> Are there applications that send markup even when server is not
> advertising that it supports "body-markup"? If yes, it would be good to
> file bugs against those instead of trying to workaround them like
> this. The specification does say "must be stripped clientside." so it
> seems really wrong to strip it server-side.

I know two:

appname: Pidgin
summary: username says:
body: test &lt;&gt;&amp;

This is actually not pidgin but the plugin "Libnotify Popups 0.14"

appname: Turpial
summary: Timeline (1 tweet)
body: <b>@username</b> Some random tweet


I agree with you (even though I have not reported bugs, shame on me), but
until this is fixed at least in the most common applications, I'd like continue
stripping those server-side.

I can change the verbose mode of dunst so that it writes something like
I pasted above to the terminal in order to identify non-conformant clients.