Bugs e-mails and R-devel [Was: Bug filed on unzip() function]
On Dec 24, 2010, at 12:22 AM, Marc Schwartz wrote:
On Dec 23, 2010, at 8:24 PM, Simon Urbanek wrote:
On Dec 23, 2010, at 7:44 AM, Uwe Ligges wrote:
This message contains a good question: Is there any reason why the bug reports are no longer mailed to R-devel?
The way Bugzilla works is that all parties involved in a bug get e-mails - but then they get all of them including all updates of the status, replies etc. One way to get involved is to be the assignee for a bug and most bugs have R-core as the assignee so that's where it goes. Although we could add R-devel on the CC list it would mean that *every* change to a bug will result in a message and I suspect R-devel subscribers would not be quite happy about that. I don't know of any provision that would make it possible to broadcast the initial report only. Moreover, doing so on R-devel would be somewhat problematic, because people might reply to all and thus some correspondence would still land on R-devel whereas replies via website would not - and that could lead to a serious confusion.
I'd appreciate to get a notice what is going on in the bug repository without having to look on those web pages.
I could add you to the CC list of any (or all) components - that's one way (it could be interesting to see how it works traffic-wise). Another would be to have a dedicated list for the bug traffic (R-bugs is not a list). Or, as I said, we could put R-devel on the CC list for all components. I wouldn't mind doing so, but I'm not sure what the R-devel readership would say... Comments are welcome. Cheers, Simon
I don't know what the volume of traffic would be from Bugzilla these days versus what it used to be from Jitterbug. One of the issues with Jitterbug and the cc'ing of bug reports and comments to R-devel, is that the e-mails would frequently come from the participants in the bug report who were not subscribers to R-devel. That required that the R-devel moderators manually approve those e-mails, which added overhead. In fact, since moving to Bugzilla, the volume of manual approvals on R-devel has declined notably since those e-mails are no longer mirrored.
That is an interesting point and confirms my feeling that the dual-mode approach has serious implications.
There is not an easy way to interact with Bugzilla via e-mail as there was with Jitterbug. The last time that I looked into this during the transition, it would require e-mails with a very specific formatting and name-value pair style entries in the message body, which could then be parsed by Bugzilla for inclusion into the underlying database. So one could not just reply to a Bugzilla bug report or comment with a free form e-mail as could be done with Jitterbug.
We work around that for R-bugs by injecting the comments directly into the bugzilla database. The rationale is that no extra e-mail notification is needed since the e-mail (hopefully) went to all parties involved so bypassing bugzilla for the update is fine. So far it seemed to work just fine. (The only additional service I was thinking of would be to allow the change of status by e-mail - using some define keyword/phrase - so you don't have to go back to the website to close a bug).
If an e-mail list mirror is desired, I would vote for a separate READ-ONLY list that folks could subscribe to and/or perhaps have an RSS feed that could be followed for updates. Making the list read-only would obviate situations where somebody replied to a bug report and/or comment via e-mail, where that reply would of course not make it into the Bugzilla repo thread, resulting in a loss of information.
Maybe the reply-to could be R-bugs which would solve the reply issue, but the original issue of non-registered users replying would still remain with even bigger consequences (the replies would not even go to bugzilla). However, I could generate bounce e-mails for those, notifying the sender that he is not registered and thus his post will be discarded - not sure if that helps, though (and it may lead to issues with spammers getting replies). Also it would increase the traffic on R-bugs which would make manual screening (which is what I do at the moment for people that try to e-mail new reports to R-bugs) almost impossible.
With Bugzilla, the results of search queries generate an RSS feed link at the bottom of the query results page (see the "Feed" link), which can be subscribed to using one's favorite RSS reader. That would be one way of keeping track of new/open bug reports.
That sounds like a good idea to me - especially since it's there already ;).
One could, if desired, create custom queries in Bugzilla using the Advanced Search functionality and then use the resultant RSS feed link to keep track of updates to the particular query result set. Also, I don't know what the typical response time has been on Bugzilla once a bug report is filed. Perhaps something could be noted there so that bug reporters might have some expectation that a comment/reply might be forthcoming within X days of filing. After that time frame, some recommended form of follow up communication could take place as a tickler/reminder of sorts.
This is happening, but only to the assignees, so currently on R-core or to individuals. Thanks for the comments, Simon
That's my $0.02. Regards, Marc Schwartz
On 21.12.2010 18:50, Ken Williams wrote:
Hi, A few days ago I filed a bug report on the unzip() function: https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=14462 I haven't gotten any comments yet, so I thought I'd ask for comments here. I also see on the description of R-devel that the list "also receives all (filtered, i.e. non-spam!) bug reports from R-bugs", but I don't see it here. Eventually I would like to help unzip() gain large-file support, such as is offered by http://info-zip.org/UnZip.html version 6.0. A corresponding zip() function would be nice too. Thanks. -Ken