1150179 - index, tools: Add deprecated bit to namespace entries

Status: RESOLVED WONTFIX

(Taskcluster :: Services, defect)

Description

[Jonas Finnemann Jensen (:jonasfj)]

So when we index a task like: ns1.ns2.ns3
We create:
  task entry:       ns1.ns2.ns3
  namespace entry:  ns1.ns2
  namespace entry:  ns1

If we added a boolean to namespace entries, so they could be marked deprecated,
that would be really cool. It's strictly a UI thing... but it would be nice not
to have old index patterns show up, while we wait for the data to expire.

------
What do people think? Good/bad idea? Do we really need this?

Comment 1

[Jonas Finnemann Jensen (:jonasfj)]

We should also add:
 - description

Comment 2

[Michael Shal [:mshal]]

garndt, is this something we could get TC to support sometime in Q4? The UI for the artifact & index browsers is currently a bit confusing without it. For various reasons we have multiple stale routes in things like gecko.v2.mozilla-central.latest (eg: linux32-dbg vs linux-debug, linux64-asan vs linux64-asan-opt, linux64-pgo-opt vs linux64-pgo, etc), and some junk or soon-to-be-deprecated routes in the top-level (ffledgling, index, buildbot, etc). It would be great if we could hide those in the UI.

Comment 3

[Greg Arndt [:garndt]]

I've added it to the backlog that we'll use for planning the quarter.  We'll definitely talk about it to find out what the right solution is.

Comment 5

[Pete Moore [:pmoore][:pete]]

Greg, this cost several people some time in bug 1361205 - I'm wondering if we should bump the priority on this one?

Jonas, is it a big piece of work?

For anyone interested in looking into it, the code for the taskcluster index service is here:

  * https://github.com/taskcluster/taskcluster-index

And the code for the Index Browser / Indexed Artifact Browser is here:

  * https://github.com/taskcluster/taskcluster-tools/tree/master/src/index

Comment 6

[Greg Arndt [:garndt]]

From the sounds of it, it doesn't appear to be a huge amount of changes, but I'll let Jonas weigh in.  It's probably small enough that it would be a good mentored bug.


One thing to note though is that this is only useful if someone goes through and cleans up old namespaces and marks them as deprecated intentionally.  For bug 1361205 it doesn't sound like anyone knew a 100% that this namespace was no longer used on mozilla-beta before digging into the bug.  It's definitely good for others after discovering this though and marking it deprecated.

I'm in favor of it! we can't have old namespaces being kept around for the end of time, things will get rather messy and confusing.

Comment 7

[Dustin J. Mitchell [:dustin] (he/him)]

I don't think we even have expiration running yet, right?  Shouldn't we do that first?

Comment 8

[Jonas Finnemann Jensen (:jonasfj)]

Neither expiration or deprecation is done, both should be fairly small things to do.

The only thing is that we really ought to be careful when we do implement expiration :)

Comment 9

[Justin Wood (:Callek)]

Is this still on someones radar, priority?

Comment 10

[Dustin J. Mitchell [:dustin] (he/him)]

Expiration is working now, and has eliminated a lot of those "old" namespaces.  The index explorer looks a lot better now.

To be honest, in my experience with hackers poking around at something and saying "ah, this thing looks like what I want", no amount of documentation, CAPITAL LETTERS, or big red warnings will prevent them from using the thing anyway.  So I have a hard time seeing a "deprecated" bit on a namespace having any measureable effect beyond that of the exiting expiration.