Identifying series

Hi folks,  OK I think I now understand how volumes fit in, and why the statement about how to uniquely identify an issue *within a series* is there on the ComicIssue page.  And that it will mostly work, except in cases where publishers behave pathologically (duplicate issue numbers- it's rare but it happens).
  The much harder question is how to uniquely identify a series, which you need to do before you can globally uniquely identify an issue.  This is where all of that stuff I was saying about the start year and the format (magazine, hardcover, paperback, etc.) comes into play.
  But there's another thorny questions, which is "what is a series?"
SHORT VERSION:
  The GCD has tended to split series into small units for maximum precision, and then provide ways to link those small series up into larger series in various ways.  This allows for different users to have different definitions of "series" depending on their use case.  But if a user wants one of the looser definitions of series, we don't have any sort of unique ID that corresponds to that.  Evolutions in the definition of a series also tend to cause the series IDs to be a little less stable than issue IDs in general.

LONG VERSION:
  A simple example is the original X-Men series, which in GCD parlance is "The X-Men (Marvel, 1963 series)".  This changed title in the 80s to Uncanny X-Men, or more precisely, "The Uncanny X-Men (Marvel, 1981 series)".  As a collector, I file these together as one series, but they are two series in the GCD.  However, the database includes a link between them that tells us that Uncanny continues the numbering of the original series, with only a "minor" name change.  So this lets us look at either both individual segments of the series, or we can say that we want to ignore minor name changes and look at that larger series.
  The types of links we currently support are:publisher change:  Fairly obvious, at least once you understand the GCD's model for publishersminor name change: At least one significant word in the title is present in both names.major name change: No significant words are present in both names.
"Insignificant words" includes things like leading articles and stuff like "Comics" or "Funnies".  "Feature Funnies" to "Feature Comics" is a minor change, because "Feature" is significant.  But if "Feature Funnies" had changed to "Awesome Funnies", that would be a major change.  There's some gray area here.
An example of a series break due to publisher change would be "Blackhawk" being bought and continued by DC when Quality shut down.
  We are discussing other possible series links, such as the sort of merges that are common in UK titles.  Here's an example where the mergers are explained in the notes:  http://www.comics.org/series/63024/
  You also get pathological weirdness like a series' numbering sequence being continued under two different names at different times:  http://www.comics.org/series/177/
  Anyway... the point is, depending on your criteria, there may be many different notions of what a series is, which makes it more complicated to identify series uniquely.  As a further example, comicbookdb takes a much more expansive view of series- for X-Men/Uncanny X-Men, not only do they combine the series over the change in title, they also include all of the related annuals and collections.  I've made arguments in the past for associating things like annual series with their source series, but I doubt the GCD would ever consider those to be the same series.  There are too many situations where the correlations aren't as clear, for one thing.
  How I envision this as working in the search part of the API is that, when asking for a series, you can specify what sort of links you want to bridge.  The most common use cases are:
* I want the smallest possible match* I want the series to ignore all minor name changes and publisher changes (this is how readers often think of series)* I want the series to ignore all name changes (this is usually about the relationship between U.S. golden age publishers and U.S. second class mail postal regulations)
In that example I linked to for Daring Mystery Comics, if you searched with the "ignore all name changes" option, you would get two results:* A composite of Daring Mystery Comics + Comedy Comics* A composite of Daring Mystery Comics + Daring Comics(probably plus other junk, there's at least one other thing called Daring Mystery floating around)
Right now, the site's search features don't do this sort of composite series thing, but it will be present in some form in the API.  Maybe returning a series object that is a composite, or maybe just returning ordered lists of series.  I haven't thought that part through yet.  But depending on the use case, an API client may be trying to uniquely identify a "series" that the GCD thinks of as a composite, and therefore does not have a single ID for.  It's not unlike the issue around how to uniquely identify a general issue as opposed to a specific variant, when all of the IDs are attached to specific variants.

thanks,-henry