A related question, maybe, is how does one delete such a publication? I think you have to edit every object with a reference in order to be able to delete all of the objects.

In terms of xml and xslt processor, I think there is little if any downside.

Although it is possible to conref from one topic to another, and create circular references, doing a lot of it can result in "spaghetti code."  This spaghetti code can at some point down the road, prevent you from being able to release a publication, because something in that publication depends on some other thing being in a released state, but that thing can't be released because it conrefs something that isn't released, and so on. Eventually, someone will want to change an item that is conref'd.  Then what?  How will they know what will be affected by that change?  How to they know it's been conref'd?

So, when you have a small number of items to keep track of, you might be able to manage it, but once you start scaling up and getting multiple writers, the complexity gets you.

There are some existing posts out there that discuss best practices for reuse, which suggest keeping a set of source topics in a particular part of the repository, and only conref from those specific topics.  For example, you might have a topic full of common alerts, and all your other topics grab the alerts from those topics. See  DITA Best Practices (Laura Bellamy, Michelle Carey, Jenifer Schlotfeldt), and for a quick illustration of what I'm talking about, see https://www.ditawriter.com/specializing-conref-target-topics-to-avoid-spaghetti-code/

Thank you, Trish. I agree completely: What you've described are the best practices for reuse in DITA.

In my case, Topic A is (for example) the enable macsec command description, Topic B is the disable macsec command description, and they reference each other. No spaghetti. However, as you point out, the potential for mischief is increased, for example if a writer edited one of the topics and added a reference to an object that was outside the publication.

Great discussion! Thanks for starting it, Larry, and thanks to Trish for describing the general best practice in DITA (helpful to a wider audience).

One of the great things around having tools to shape on top of the standard, is that you can reinforce best practices and somewhat discourage practices that could get the unwary into trouble. So we do recommend that customers exclusively use library topic types as sources for conrefs. And hence the warning message (more of a nudge I hope than a blaring klaxon!)

Larry, in this case would it work to house the common passage that's reused between those topics in a dedicated library topic? I know there are always exceptions but that is the "by the book" recommendation here I'd say.

P.S. – you don't have to but I always used to get the writing team adding the library topics to the Resources "folder" at the bottom of the publication, as a clear, quick place to see not only variable sources but conref sources too.

I would say that conrefs between topics is a dangerous path. We use a LOT of conrefs, but our house rules specify that we shall only do this FROM library topics TO topics, and never both ways!

I do see this message, but only occasionally, even with publications using more than 5000 objects. It has actually never bothered me.