You can clone with
No one assigned
Try::Tiny was not indexed #234 and other files are missing, while they are present on SCO, because the author hasn't included them in the MANIFEST.
Question is: should we be trying to handle this gracefully, rather than just not indexing the file?
Also, from #112:
links on main page broken as a resulthttp://beta.metacpan.org/module/Forward::Routes
MANIFEST is afaik not required as part of a valid release (I might be wrong).
Well, if it made it past PAUSE, it's either valid or there's a bug in PAUSE.
A MANIFEST is indeed not required. In fact, nothing really is required in PAUSE - you can upload any kind of file, archive or not.
I believe this issue boils down to whether or not we want to maintain a set of heuristics to prevent things that have been intended to be full distributions, but which are obviously broken.
Given how many differences exist in the style people create their distributions and how much leeway the CPAN toolchain gives authors, I think we don't.
Index everything. Flag (in some way) anything that's "wrong", but still index it.
For what it's worth, this isn't what the issue was with Try::Tiny. The issue was that the broken dist contained no files at all (the manifest was empty, and so "make dist" didn't include any files in the built dist). search.cpan.org saw that there were nothing in the dist that was actually indexed, and so preserved 0.09 as "latest", while metacpan just saw that a dist named "Try-Tiny-0.10.tar.gz" existed, and assumed that 0.10 should be the latest, even though "cpanm Try::Tiny" would still install 0.09. I think there should at least be some heuristic about checking to see if the dist contains any indexed modules.
what sco does (well, what I think it does), is parse the 02packages.details file and using that data decides which release and file is latest to a given module name. I'm in the process of reworking the "latest" script to do the same.
That sounds like the right way to go to me.
I think this has been fixed now, since I rewrote the 'latest' script. It uses the reliable data provided by PAUSE in the 02packages.details.txt file.
The Try::Tiny problem would have been fixed by this as well. Unfortunately @doy released a new version :)