In Container view, render uncontained processes as distinct pseudonodes

[tomwilkie]

[peterbourgon]

"Not contained" is much less elegant 😛

http://dictionary.reference.com/browse/uncontained

[tomwilkie]

Given the discussion re: pseudo nodes ("everything in a topology should be
the same kind") I'm inclined to just remove uncontainer. Especially when
we have a real containers topology with all containers in it.

On Thu, May 28, 2015 at 10:30 AM, Peter Bourgon notifications@github.com
wrote:

"Not contained" is much less elegant [image: 😛]

—
Reply to this email directly or view it on GitHub
#127 (comment).

[peterbourgon]

I don't understand the motivation. The proposal is to remove information from the UI, which undermines a lot of our user stories. Concretely: you want to see if a container is talking to more than just other containers. That single bit of information tells you a lot — it can identify misconfiguration, false assumptions you've made about your infrastructure, even identify backdoors or broken third-party software. At the moment I can't see how a topology with 1 less node is worth losing that value. Can you convince me?

[tomwilkie]

A large part of good UX is consistency; making everything in a single topology the same type make explaining what it is your showing to users easier (ideally to the extend where you don't need to explain). Exceptions (they are all containers, except that node represents incoming stuff from the internet, and that node represents stuff not in a container) make the UI less intuitive.

I agree the use cases you've highlighted are important ones; perhaps there is an alternative? Perhaps uncontained workloads should be pseudo nodes (not all grouped into one, perhaps per-node), and perhaps we should revisit your suggestion of rendering pseudo nodes differently (smaller, greyed out...).

[peterbourgon]

Exceptions . . . make the UI less intuitive.

I agree, but I think only a little bit less intuitive, and at a cost of a lot of loss of information. In this instance, I think the calculus doesn't work. That said, I'd be happy to explode the single Uncontained node into multiple per-host nodes. Shall we re-scope this ticket to that?

And I think we should definitely render all pseudo-nodes differently. Filed #131.

[peterbourgon]

Are we OK closing this in favor of #131?

[tomwilkie]

I think there are a few things that need doing before we can close that
aren't covered by #131:

• break out uncontained into individual, per container/host pseudo nodes
• bring back 'normal' pseudo nodes for this view.

I'm kinda waiting in your big PR before I have a crack at this.

On Wednesday, 3 June 2015, Peter Bourgon notifications@github.com wrote:

Are we OK closing this in favor of #131
#131?

—
Reply to this email directly or view it on GitHub
#127 (comment).