taskSavane - Tasks: task #1074, mail notification content...

 
 
Show feedback again

You are not allowed to post comments on this tracker with your current authentification level.

task #1074: mail notification content improvement

Submitted by:  Mathieu Roy <yeupou>
Submitted on:  Tue 23 Nov 2004 09:29:11 AM UTC  
 
Should Start On: Thu 18 Nov 2004 12:00:00 AM UTCShould be Finished on: Fri 18 Nov 2005 12:00:00 AM UTC
Category: Web FrontendStatus: Done
Priority: 3 - NormalPlanned Release: 
Assigned to: Mathieu Roy <yeupou>Open/Closed: Closed
Privacy: PublicFor/By: None

(Jump to the original submission Jump to the original submission)

Sun 13 Mar 2005 10:19:04 AM UTC, comment #24:

"As for <>, I think I never saw a tracker item URL split into several lines, so maybe it is not really needed either "

In case they're a white space in the URL, it will make the different (at least with gnus).

I'm frankly happy with the current notifications we have. Easy to read, fast to get to the point. I close this item but people is free to submit bug if there's a problem.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Sat 22 Jan 2005 03:08:42 PM UTC, comment #23:

"All mail clients suck. This one just sucks less." http://www.mutt.org :)

Anyway, I just remembered that there is a command to easily browse a list of URLs from the mail and launch them, so this is not really needed.

As for <>, I think I never saw a tracker item URL split into several lines, so maybe it is not really needed either

Statu quo, I guess.

Sylvain Beucler <beuc>
Project Administrator
Sat 22 Jan 2005 09:52:08 AM UTC, comment #22:

Hum, I added the <> because many text mailers (like gnus) treat the content between these as links (as a complete string, which means <http://gnag.gag/ad adadda.html> will be completely handled, unlike http://gnag.gag/ad adadda.html that would make a broken link).

I do not know if there is a standard about that. It is hard to make a decision that will help you but potentially be a pain to others.

That being said: is there no way to click (or push enter) on a link with mutt? Hu, how can someone use such tool :P

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Fri 21 Jan 2005 09:53:03 PM UTC, comment #21:

Even less info? Well yes.

My TZ is (un)configured to GMT atm which can explain the conflict.

Last, could you consider removing '<>' around URLs? I'm a mutt user, and if those weren't added, I could triple-click in the terminal to copy the URL to the clip-board (and then middle-click in Firefox). Right now, I have to do a precise selection with the mouse, that's terrible :p

Sylvain Beucler <beuc>
Project Administrator
Wed 19 Jan 2005 09:53:05 PM UTC, comment #20:

Even less information? Arg, it will be really hard to know what it is about. We already seriously removed a lot so far, do you think we should go even further :] ?

"I meant that having to use a browser to send a notification may not be necessary in the future (ie if we get such a mail interface), and that should be taken into account when making long-term decisions."

You are right. But we cannot (okey, we did that in the past) consider ok to send each time a mail with as many information as the item webpage. The item webpage must be exhaustive because we need a place like that at it fits well this purpose.
Indeed, the day a mail interface is implemented, it should provide a way to get exhaustive information also. But the mail is currently used mainly as notification, and will still be used as notification; as such it has constraint the webpage does not really have.

---

Side note: this time, theoretically, we should not get timezone conflicts (you are not in Europe/Paris timezone?)

Should Start On: Thu 11/18/2004 at 23:00 => Thu 11/18/2004 at 00:00
Should be Finished on: Fri 11/18/2005 at 23:00 => Fri 11/18/2005 at 00:00

since the relevant bug is hopefully fixed.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Wed 19 Jan 2005 09:28:59 PM UTC, comment #19:

To keep it clear (this time), I suggested you replace:

"<blank line>
Changes:

Assigned to: None -> Beuc
Status: None -> Won't do

_____________________________________________________

Follow-up Comment #17:

I am sorry but I doubt you can convince me that a comment is more important than the field that defines the status, priority (etc) of an item.
[...]"

by

" Assigned to: None -> Beuc
Status: None -> Won't do

I am sorry but I doubt you can convince me that a comment is more important than the field that defines the status, priority (etc) of an item.
[...]"

> > I do not dispear to see / code a mail interface to the
> > trackers i> the>future.
> But such interface could be an additionnal interface, not a
> replacement.


(By "mail interface", btw, I meant that replying to a notification will auto-add the mail reply's content to the tracker.)

I meant that having to use a browser to send a notification may not be necessary in the future (ie if we get such a mail interface), and that should be taken into account when making long-term decisions.

Sylvain Beucler <beuc>
Project Administrator
Wed 19 Jan 2005 08:58:56 PM UTC, comment #18:

Now ", tracker #nnnn (project unixname):" is added to the first subtitle.

I hesitated with ", tracker #nnnn, project unixname:" but was unable to decide so I picked the first.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Wed 19 Jan 2005 08:37:02 PM UTC, comment #17:

I am sorry but I doubt you can convince me that a comment is more important than the field that defines the status, priority (etc) of an item. Comments are technically here to justify changes made on these fields. These fields are what matter most. We are talking about an issue tracker.

Does rfc822 defines issues trackers notifications? What matters is not to print data, it is also to make it understandable, even by non-geeks. I'm not sure we can find something easier to understand than the current

$field: $old_value -> $new_value

What surprises you is maybe the formatting. But from what I know, it is more a nuisance to a user to deal with stuff like

$field: $oldvalue -> $new_value
$thisfieldislonger: $huthisislongtoo -> $short
$v: $a -> $a

"would not remind me much about this task within a few weeks."

I guess that when the title does not speak for itself, one have to follow the link (which is nowadays quite trivial). Just like if someone was receiving a mail on a mailing-list.

"I do not dispear to see / code a mail interface to the trackers i> the>future."

But such interface could be an additionnal interface, not a replacement.

Anyway, sending full information in each notification defeat the concept of "notification". I guess a mail interface would provide a way to get the full details of an item, given a specific command.

"The drawback is that web users may get bored to see a lot of quoting in the
comments; in this case, maybe we should provide a way to filter such quoting
for web users. If we get super-clever when comparing the quoted texts, we
could even provide an auto-threaded wiki-discussion-style way to present the
follow ups, but that can wait :)"

It's getting complicated. :)

But currently, I'm quite satisfied with the content of the notifications (minus the bug "ven 19.11.2004 à 06:00 -> ven 19.11.2004 à 00:00"), I just had to take care at first to read the title field, something I forgot. The previous notification was too dense that I was anyway usually looking at the item on the web interface anytime I needed context.

So:
- me alive, "changes" will stay before "comments" :))
- anything that could remind the user would be good
- somewhere, the project name is missing

I think I'm going (right) to make a test with the "Changes:" and "Follow-up Comment #16:" section title, adding ", project $name, bug $nnn:" to the first section shown on the mail. That would address the 3rd point.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Wed 19 Jan 2005 08:05:52 PM UTC, comment #16:

The follow-up is IMHO the most relevant part of a notification. So the solution I have in mind is to write the follow-up at the top, and then, put the other infos.

If people like the ones working at CERN do need the changes to be placed before the follow-up, then consider removing the "changes:" title, as well as the "follow-up comment #xx" one, and the "---" separator; to sum up, use RFC822-style organisation:

change1: value1
change2: value2
<blank line>
follow-up

and then, all the additional information needed, structured whatever way is needed. Using this RFC822-style, it is obvious which part describe the field changes and which part describe the follow-up, so no additional presentation text is needed.

Regarding the history, quoting only the latest comment is now necessarily good. In this very case

> Hum, that's correct, the original subject does not permit to deduce it.

would not remind me much about this task within a few weeks.

> But they are more or less forced to [use a browser] anyway, if they want to further comment the issue.


I do not dispear to see / code a mail interface to the trackers in the future.

Note that if we get a mail interface, and maybe a quoting facility for the web, then it would be up to the users to select the history they want to include in the notifications. Just like we do when discussing in the mailing list. Items would be bigger in the database (because of all the duplicated quoted history), but posting the history in notifications would then be rather useless and that would solve the problem of including history.

The drawback is that web users may get bored to see a lot of quoting in the comments; in this case, maybe we should provide a way to filter such quoting for web users. If we get super-clever when comparing the quoted texts, we could even provide an auto-threaded wiki-discussion-style way to present the follow ups, but that can wait :)

Sylvain Beucler <beuc>
Project Administrator
Wed 19 Jan 2005 07:43:28 PM UTC, comment #15:

Definitely, the message are more "lisible" now. I like it very much.

If Savane implements the scheme, described in task #1173, then all you need to keep track of context is to use thread-capable mailer. When user asks Savane for previous messages, Savane inserts correct Messages-ID and In-Reply-To fields in them and sends them to the user. Then user's mailer constructs the threads, and the user will have all the context: each change in its proper message, all of them in one thread.

The meta information could be MIME'ed and placed into separate MIME-part. This way, MIME-capable mailers will be able to show different parts of message differently (or even show them as a tree).

P.S. Would be nice to have threaded comments (not flat ones, as in current system).

DIG <dig>
Wed 19 Jan 2005 06:47:41 PM UTC, comment #14:

Hum, that's correct, the original subject does not permit to deduce it.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Wed 19 Jan 2005 06:07:10 PM UTC, comment #13:

I think simple follow-up on this tracker comments lack the presence of the group name (in our case, 'Savane'). Since I recieve double notifications (one via the ML, one via direct mail), I can only link to the group when reading the ML one thru its [bla-dev] header in the subject.

Vincent Caron <zerodeux>
Project Member
Wed 19 Jan 2005 04:44:55 PM UTC, comment #12:

Note: normally, in a mail, the subject of the mail should be enough as regard of reminding the context.

So I guess it is also a matter of updating correctly the item title field. Or maybe comments should have titles?

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Wed 19 Jan 2005 04:41:20 PM UTC, comment #11:

> I don't care about most of the metadata fields, ie I only >care about whether the item is open or closed (plus why it >was closed), and who is working on it.


But that would not be acceptable for several people that I know expecting from the notification to be exhaustive about the changes. Reminding context is questionnable. Mentioning change is a must, not really an option.
(remember that there are project out there, especially at CERN, that use the trackers in a very precise workflow)

>The disappearance of the history is very inconvenient at >the moment, because Mathieu is running through bug reports >that I don't remember of, which made me load the tracker >www page each time.
>
>I think I would vote for a solution where you basically >have only the last reply at the top of the mail (no >separator, no title, no heading blank line), then the >notification as in the current trunk. The history part >should be in a consistent order; previously you had the >original submission (oldest part), then the follow-ups >from the newest to the oldest, which is sometimes not easy >to read.


Considering that removing the changes section is not an option, we cannot get rid of all separators.

Having the last reply at the top of the mail basically would defeat your own request about having the relevant information in the very first lines of the mail.

I found reasonnable the request to keep the history in the web interface, off the notification. So yes, people are likely to need to use their browser if they need it. But they are more or less forced to do that anyway, if they want to further comment the issue. Having the history force us to use many many more kind of separators and most of the time end up in something not very lisible (text only is somewhat limited in this regard).

In think the only way to see these notification is to treat them like mails.
One option would be to add automatically in forms the previous message, prefixed by >.

> Also, did you consider templatizing the notifications?


Yes, and was against.

- If it is about have site-specific templates : we loose the interest of a tool like Savane, which is providing a way to do things (if each installation of savane requires 3 month to get used to, it is not fine - someone who knows savane should feel familiar on every savane installation).
- If it is about providing user-specific templates, it is likely to exhaust the www server just as like the mail server, just to serve a few mails (because it would have to generate one new mail per person participating to a discussion) and... exhaust developers with a code unbearable from a maintainance point of view.

So there's a question here: should we put back a part of the history (which would drastically degrade lisilibity but enhance context understanding)? Should we by default propose the content of the previous post, just like a mailer do when you do a reply?

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Wed 19 Jan 2005 04:19:44 PM UTC, comment #10:

As far as I am concerned, I find it difficult to get the essential information in the previous version: I often 'miss' one-line comments among all the separators and blank lines. It is good when you have the relevant information in the very first lines of the mail (no scrolling needed).

I don't care about most of the metadata fields, ie I only care about whether the item is open or closed (plus why it was closed), and who is working on it.

The disappearance of the history is very inconvenient at the moment, because Mathieu is running through bug reports that I don't remember of, which made me load the tracker www page each time.

I think I would vote for a solution where you basically have only the last reply at the top of the mail (no separator, no title, no heading blank line), then the notification as in the current trunk. The history part should be in a consistent order; previously you had the original submission (oldest part), then the follow-ups from the newest to the oldest, which is sometimes not easy to read.

It would also be possible not to include all the comments, since it tend to generate message that are bigger than the Mailman default size of digest (and you receive 1-mail digests at savane-dev ;)).

Also, did you consider templatizing the notifications?

Sylvain Beucler <beuc>
Project Administrator
Wed 19 Jan 2005 03:34:59 PM UTC, comment #9:

I'm a bit afraid of the lost of context too, we'll see how users react to it. That was the thing that made me reluctant to make heavy changes like this in the first place.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Wed 19 Jan 2005 02:46:41 PM UTC, comment #8:

This is definitively more easy to read, since most of the time only new data is sent. On the other side, we're loosing some context info. However it's very easy to follow the ticket link, and have it all in a nice Savane-formated page. I like this compromise a lot, congrats for those efforts !

Vincent Caron <zerodeux>
Project Member
Wed 19 Jan 2005 05:40:16 AM UTC, comment #7:

Done. I submitted new task #1173.

P.S. By the way, it's DIG (these letters are initials).

DIG <dig>
Wed 19 Jan 2005 05:16:27 AM UTC, comment #6:

Sure, I will do it.

DIG <dig>
Tue 18 Jan 2005 10:50:08 PM UTC, comment #5:

I made some changes to reduce the amount of data sent, assuming we wont mention in the mail content what can be guessed/known by reading the mail header.

I upgraded Gna! to the relevant branch. Please, play with it and post your comments so we can improve this, make sure we do not miss anything important.
Do not hesitate to make proposals.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Tue 18 Jan 2005 09:41:52 PM UTC, comment #4:

Hello Dig,

Can you submit a specific task for your proposal?

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Mon 03 Jan 2005 07:59:18 AM UTC, comment #3:

At this moment the user can subscribe himself to receive a CC mail notifications about the updates. I think it would be nice to give him the possibility to sent himself previous discussion, something like "Send me previous discussion" button or check box near the field "Add an email address that should receive carbon copies of updates of this item". This way the user will be able to follow the updates and to have previous, nicely threaded discussions in his mail box.
So there will be no need to reiterate the same content over and over again.

DIG <dig>
Sun 02 Jan 2005 01:04:37 PM UTC, comment #2:

I think I'm going soon to try to improve the mail notif by removing the overview content. Someone also (dont remember who) suggested to skip things like "posted by" as we can assume the information is already in the from headers. It makes sense to me.
Without the overview section, we gain a lot of flexibility. Because it means, more or less, than most sections just get removed.

I intend to make a branch soon for this cosmetics changes.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.
Tue 23 Nov 2004 10:57:17 AM UTC, comment #1:

BTW, the resulting notification is there: https://mail.gna.org/public/savane-dev/2004-11/msg00251.html

There's a thread on savane-dev there: https://mail.gna.org/public/savane-dev/2004-11/msg00248.html

And I forgot to mention: yes, the email notifications have recently become much slicker, thanks ! I just realized I started to comment on them exactly because they were improved, and jumped on the bandwagon...

Vincent Caron <zerodeux>
Project Member
Tue 23 Nov 2004 09:29:11 AM UTC, original submission:

Efforts have been made in the latest development. The result is far from perfection but lisibility is probably now better.

I expect more improvement to be made for notification in the future. But that a part hard to improve. Most of the time, the complainers have a very specific vision of what would be the perfect notification while they do not have a clear idea of what the notification could contain. For instance, persons that usually tracks issues where item update only contain 2 lines followup comments tend to think the separators to be overly line consuming; but if they were receiving followup comments of 50 lines including bits of code, with 6 fields being updated, they would clearly understand the waste of one line as separator that greatly enhance lisibility.

My personal stance is that what should matters most is not the amount of lines used by the lisibility in itself. Tests should always be made with item with long comments and many field changes.
Having less information is usually convenient for regular users, since they know exactly where they should be looking. But the interface should be newcomers-proof as much as possible.
For instance, a first paragraph saying the mail is a notification about this item on this tracker on this installation is probably useless for many users. It is not for newcomers, it is not completely the case for persons that uses several instance of Savane. One could interpret the mail header to get clues about the context; but it is not sure at all.

Many users never read what is written, even when should. I tend to think that it is up to them -- it is their problem in the end if they missed something by not reading. But users that do read what is written should never miss something because we were pretentious enough to assume that everything is crystal-clear and does not require any comments.
I personally rarely read some parts of the Savane interface. I almost never read the Label of the fields (like Category, Privacy), I only click on the select boxes. Maybe it would be even more confortable to me if there was no such label. It would be possible to get a form even smaller, with less lines wasted. That would be me just taking into account my own usage of Savane and what would be very efficient for me.
It is probably the way Emacs was designed. Emacs is powerful, I use it everyday. But it tooks me weeks just to understand how to do the most basic things.
Savane should not be an Emacs, should not be a tool designed for computers experts.
If we can make happy both computers experts and the rest of the world, ok. But when we cant, we should make an effort towards the rest of the world.

These two points (items will plenty of info, problem of accessibility to end user with little computing knowledge)
are why the latest changes made to notification did not reduced verbosity and separators line consumption.

I'd like everybody that want to work on notification content improvement keep these two points in mind. It does not means that separators and verbosity should never be changed. It means that when we'll speak about changing it, we'll do it with in mind that the result will be as easily understandable to most users.

-------------------------------

That said, I wonder if the section "Overview" is justified in mail changes notifications. Maybe it should be sent only in the first mail and the update notification should only mention what changed.
It would greatly reduce mail size and would allows us to change things more easily, as we would no longer need big OVERVIEW/CHANGES separators.

That's, I think, a starting point for future improvements. But I have no clue whether the overview section is felt as important to users or not. I know that a few said it was useless. But a few complainers is not enough to know what the others thinks.

Mathieu Roy <yeupou>
Project AdministratorIn charge of this item.

 

No files currently attached

 

Digest:
   task dependencies.

Items that depend on this one: None found

 

Carbon-Copy List
  • -unavailable- added by beuc (Updated the item)
  • -unavailable- added by dig (Updated the item)
  • -unavailable- added by zerodeux (Posted a comment)
  • -unavailable- added by yeupou (Submitted the item)
  •  

    Do you think this task is very important?
    If so, you can click here to add your encouragement to it.
    This task has 0 encouragements so far.

    Only logged-in users can vote.

     

    Please enter the title of George Orwell's famous dystopian book (it's a date):

     

     

    Follow 25 latest changes.

    Date Changed By Updated Field Previous Value => Replaced By
    Sun 13 Mar 2005 10:19:04 AM UTCyeupouStatusGLOBALS=>Done
      Assigned toNone=>yeupou
      Open/ClosedOpen=>Closed
    Wed 19 Jan 2005 09:28:59 PM UTCbeucShould Start OnThu 18 Nov 2004 11:00:00 PM UTC=>Thu 18 Nov 2004 12:00:00 AM UTC
      Should be Finished onFri 18 Nov 2005 11:00:00 PM UTC=>Fri 18 Nov 2005 12:00:00 AM UTC
    Wed 19 Jan 2005 08:59:24 PM UTCyeupouPriority1 - Later=>3 - Normal
      StatusNone=>GLOBALS
    Wed 19 Jan 2005 08:37:02 PM UTCyeupouShould Start OnFri 19 Nov 2004 12:00:00 AM UTC=>Thu 18 Nov 2004 11:00:00 PM UTC
      Should be Finished onSat 19 Nov 2005 12:00:00 AM UTC=>Fri 18 Nov 2005 11:00:00 PM UTC
    Wed 19 Jan 2005 08:05:52 PM UTCbeucShould Start OnFri 19 Nov 2004 06:00:00 AM UTC=>Fri 19 Nov 2004 12:00:00 AM UTC
      Should be Finished onSat 19 Nov 2005 06:00:00 AM UTC=>Sat 19 Nov 2005 12:00:00 AM UTC
    Wed 19 Jan 2005 07:43:28 PM UTCdigShould Start OnFri 19 Nov 2004 11:00:00 PM UTC=>Fri 19 Nov 2004 06:00:00 AM UTC
      Should be Finished onSat 19 Nov 2005 11:00:00 PM UTC=>Sat 19 Nov 2005 06:00:00 AM UTC
    Wed 19 Jan 2005 04:41:20 PM UTCyeupouShould Start OnSat 20 Nov 2004 12:00:00 AM UTC=>Fri 19 Nov 2004 11:00:00 PM UTC
      Should be Finished onSun 20 Nov 2005 12:00:00 AM UTC=>Sat 19 Nov 2005 11:00:00 PM UTC
    Wed 19 Jan 2005 04:19:44 PM UTCbeucShould Start OnSat 20 Nov 2004 11:00:00 PM UTC=>Sat 20 Nov 2004 12:00:00 AM UTC
      Should be Finished onSun 20 Nov 2005 11:00:00 PM UTC=>Sun 20 Nov 2005 12:00:00 AM UTC
    Wed 19 Jan 2005 02:46:41 PM UTCzerodeuxShould Start OnSun 21 Nov 2004 06:00:00 AM UTC=>Sat 20 Nov 2004 11:00:00 PM UTC
      Should be Finished onMon 21 Nov 2005 06:00:00 AM UTC=>Sun 20 Nov 2005 11:00:00 PM UTC
    Wed 19 Jan 2005 05:16:27 AM UTCdigShould Start OnSun 21 Nov 2004 11:00:00 PM UTC=>Sun 21 Nov 2004 06:00:00 AM UTC
      Should be Finished onMon 21 Nov 2005 11:00:00 PM UTC=>Mon 21 Nov 2005 06:00:00 AM UTC
    Tue 18 Jan 2005 09:41:52 PM UTCyeupouShould Start OnMon 22 Nov 2004 06:00:00 AM UTC=>Sun 21 Nov 2004 11:00:00 PM UTC
      Should be Finished onTue 22 Nov 2005 06:00:00 AM UTC=>Mon 21 Nov 2005 11:00:00 PM UTC
      Dependencies-=>Depends on task #1170
    Mon 03 Jan 2005 07:59:18 AM UTCdigShould Start OnMon 22 Nov 2004 11:00:00 PM UTC=>Mon 22 Nov 2004 06:00:00 AM UTC
    Show feedback again

    Back to the top


    Powered by Savane 3.1-cleanup