Almost

I've re-integrated the linking mechanism into the new data structures so it should be working fine now and I've turned everything back on.

Unfortunately my battles with the family trees are not over. First level trees seem to be handling collabs with connectivity just fine:
Neil Young -> Crazy Horse -> The Rockets

but 2nd level trees are not. Instead of the correct:
Crosby, Still, Nash, & Young -> Neil Young -> Crazy Horse -> The Rockets

I'm getting:
C, S, N, & Y -> Crazy Horse

It seems the enemy has retreated but not surrendered. The struggle continues...

I'm sure there will be a few quirks that will pop up on various pages due to all the recent changes, so please help point them out so I can get everything back to normal. Thanks.


Kevin

Glad to see that this has all been resolved now. Another great achievement for the site. One small question, though... when we discussed fixing the trees before, I thought the idea was that we would see:

Neil Young -> Neil Young & Crazy Horse -> The Rockets

rather than:

Neil Young -> Crazy Horse -> The Rockets

Doesn't the latter imply that "Neil Young" shares membership with "Crazy Horse", i.e., that Neil Young was a member of Crazy Horse? Or would it be too difficult to get the full collab to show in the family tree like that?

While the NY & CH can be compelling, this is how I think about the situation. In my mind regarding collabs there is no such thing as a band called "Michael Jackson & Paul McCartney," that simply does not exist. While that may not have connectivity there also isn't a band called "Kurt Cobain & William S. Burroughs" either, which does have connectivity. While it could be a toss up with something like NY & CH, the solo + solo situation of collabs with connectivity defines the standard and all other situations then fall in line.

As always I reserve the right to be fickle and change my mind at any time, but that's my take on things currently.


Kevin

I'm with bgz on this one - by our rules there *is* an entity/band/thing called "Kurt Cobain & William S. Burroughs". That is the ultimately 'correct' representation of what we do with connective collabs.

Yeah, I think bgzimmer and misterpomp's views on this seem more compelling to me...

There is such an entity as "William S. Burroughs · Kurt Cobain." Our rules define it as a collaboration with connectivity, and not a band. Should it be an entity that shows up in the family tree?

Why do we have collaborations with connectivity at all? I believe the thinking is, if two previously-existing recording entities get together to record, and don't take the time to come up with a new name for themselves, then it's not really a new band, and there's no connectivity. But if any of those entities has not established an entity previously (Cobain in this example), then we connect them, since it represents the formation of something new. That, in my mind, means we ought to represent it on the tree, because it is the means of connection.

Here's something else to think about, while we're talking about this:

I think everyone agrees that there's a band called "Tom Petty And The Heartbreakers." Even though Petty has recorded solo work, we don't separate them because The Heartbreakers have never recorded anything without Petty. So it's a band through and through.

If bandtoband existed in 1970, we would have said that Neil Young And Crazy Horse was a band. Same situation: Neil Young had released material solo and with Crazy Horse at this point. It wasn't until Crazy Horse released an album on their own in 1971 that it became a collaboration. So why isn't Everybody Knows This Is Nowhere credited to a single band?

By our rules, if The Heartbreakers released an album tomorrow, Tom Petty And The Heartbreakers would cease to be a band, even for all of the previous recordings. That just feels wrong to me. I think we should only break up the entities after they've broken up themselves.

That's a good argument, ajw. If it was a band when it was released - it remains a valid band. What is your thought on a later collab after both parties have released material independently?

Well, there are two tacks you could take here:

(1) They should be separated afterwards. So, Everyone Knows This Is Nowhere should be by the band "Neil Young With Crazy Horse," but Sleeps With Angels should be by the connected collaboration "Neil Young" and "Crazy Horse."

(2) Once a band, always a band. So, both Everyone and Sleeps should both be by "Neil Young And Crazy Horse" (their connector changes frequently, it turns out, but we could use aliasing to fix that).

One one hand (1) makes sense because, after they've established themselves separately, they can decide whether or not to record with each other, and it's probably less of a band dynamic. On the other hand (2) makes sense because it's like they're getting the band back together after spending some time apart, back to an estabilshed entity. So I don't know. I lean more towards (1), but (2) wouldn't make me upset. I'm open to better arguments on either side, because neither of these really convinces me.

Beyond the existential question of whether connected collabs are considered "entities" in our system, I'm more concerned about a misleading display of information, particularly for visitors who aren't familiar with our convoluted collab rules. Currently, if I look at Art Blakey's band page, I'll see this in the family tree:

Art Blakey -> The Jazz Messengers

So was Art Blakey ever a member of The Jazz Messengers? Well, in this case, yes, because he's treated as a member for the "Horace Silver And The Jazz Messengers" album. But what about Horace Silver's band page? It shows:

Horace Silver -> The Jazz Messengers

A casual observer might assume Silver was also a member of The Jazz Messengers, which he wasn't. If, instead, we saw:

Horace Silver -> Horace Silver & The Jazz Messengers

...then there would be no confusion at all.

I'm glad that I reserved the right to change my mind given the current feeling on the topic. Right now I'm spending time really trying to reduce the code complexity as well as the memory footprint that the current Mapping / Tree functions take on the system when running (too bad everyone can't take part in that *fun* process). Once I can tighten up the system I'll take a crack at putting together the complete "NY & CH" style family trees so we can take a look and evaluate that version of family trees versus the current style. I'll let everyone know when I can get that done.

Kevin

Well, I'm always up for some code optimization.

I finally finished the code optimization and clean-up and after all that I realized that while the new data storage works in every case on the site that I know of for creating maps but there is a theoretical situation that it doesn't handle correctly. Of course, to me, that is unacceptable so I'm going to have to go in and re-do everything. Sweet.

Sorry but for now all the other requests that need my attention are going to have to be kept on hold a little while longer.


Kevin