launchpad-dev team mailing list archive

[Robert Collins <robert.collins@xxxxxxxxxxxxx>]

On Fri, Jan 28, 2011 at 3:55 AM, Gavin Panella
<gavin.panella@xxxxxxxxxxxxx> wrote:
> On 26 January 2011 18:28, Deryck Hodge <deryck.hodge@xxxxxxxxxxxxx> wrote:
> ...
>> The other source of changing comment numbers, which can still be a
>> problem today, are bug watches that sync comments.  This can change
>> the comment number on a bug page, since imported comments inlined
>> according to timeline and given comment IDs.
>
> I think imported comments should not have comment numbers.
>
> Their URLs on Launchpad should probably be based on their ID, or the
> #<num> links (obviously without a number) should point directly back
> to the comment in the bug tracker from whence they came.

But /today/, they do, right ?

>> When Robert and I spoke about this, we discussed adding imported
>> comments on the end of the bug report to better indicate when the
>> imported comments were added relative to the LP bug page.
>
> I think this would be more confusing. I think it's important to show
> messages chronologically. Imported comments do have a distinctive look
> already; perhaps they can be made more so.

Adding new comments to the end of the discussion *is* chronological
for an observer using Launchpad. It's not globally chronological, but
we know that that is an illusion :).

> In general, I agree that comment numbers should be written once. I
> don't think holes in the numbering are a problem. At least it's
> obvious there's something missing. More confusing is renumbering.

So all I'm doing is:
 - ensuring all BugMessages for a Bug have a cached index
 - only writing them once
 - assigning new indices in chronological order

Is this sufficient, or do I need more complex logic? Once done, this
can be changed by using a data migration script to implement any new
policy you might have; my primary goal is to kill with prejuidice our
top timeout.

-Rob