naming scheme for generating ResourceSync documents

[markmatney]

I've observed the following about the way this implementation names generated sitemaps (resourcelist, changelist, etc.) when ResourceSync.execute() is invoked (for a set of 50000 resources or less):

• With strategy 0 (resourcelist):
  • On first invocation, a file resourcelist_0000.xml is created.
  • Subsequent invocations modify that file and keep the same name.
• With strategy 1 (new_changelist):
  • On first invocation, a file changelist_0000.xml is created.
  • On subsequent invocations, a file changelist-index.xml is created, and points to a series of changelist_xxxx.xml files.
• With strategy 2 (inc_changelist):
  • On first invocation, a file changelist_0000.xml is created.
  • Subsequent invocations modify that file and keep the same name.

I've not tested on collections with more than 50000 resources, but per the spec, I expect there'd be a resourcelist-index.xml generated that points to a number of resourcelists.

I am thinking that whenever there is only one resourcelist or changelist for a set of resources, then it should be named simply resourcelist.xml or changelist.xml, without a _0000 suffix, on the basis that a suffix like that is misleading/unnecessary when it is the only such sitemap for a set of resources.

Thoughts?

[hariharshankar]

I am thinking that whenever there is only one resourcelist or changelist for a set of resources, then it should be named simply resourcelist.xml or changelist.xml, without a _0000 suffix, on the basis that a suffix like that is misleading/unnecessary when it is the only such sitemap for a set of resources.

The main reason for using this naming scheme is consistency. If we name the first file either resourcelist.xml or changelist.xml, then the subsequent files with name _***.xml will have no continuity with the first file and will cause confusion. The current naming scheme, I think, is very intuitive even for people with no knowledge of the protocol.

[markmatney]

I agree that that would be confusing; whenever there is more than one <capability>.xml for a set of resources, it would suffice (and be a good idea) to rename the first file to <capability>_0000.xml. However, that may never even be necessary, since it's possible that only one changelist may ever exist for a set of resources (by exclusively invoking strategy 2 for changelist generation). But, when I do see a file named changelist_0000.xml, I am inclined to expect to also see changelist_0001.xml, for example. I think a good sanity check would be to just call the singular changelist changelist.xml.