Deprecation of antecedent (old) SNOMED works

As I'm sure you've already seen, we've sent out several reminders that as of April 26, 2017 the antecedent SNOMED works will no longer be licensed. Please see here for the various reminders (we've also just reminded everyone in the latest Member Forum meeting):

http://www.snomed.org/news-articles/timetable-for-the-withdrawal-of-legacy-snomed-codes

http://www.snomed.org/snomed-ct/what-is-snomed-ct/previous-versions-of-snomed-ct

As we all know, the SNOMED CT International Edition includes an antecedent SNOMED identifier reference set, which contains SNOMED, SNOMED II, and SNOMED International (3.x) concept identifiers. This reference set will be withdrawn according to the following timeline, (which was published in 2015):
. July 2015 SNOMED CT International Edition: No change regarding antecedent SNOMED identifier reference set.
. January 2016 SNOMED CT International Edition: The last release to include the antecedent SNOMED identifiers for new SNOMED CT content.
. January 2017 SNOMED CT International Edition: The last release to contain the antecedent SNOMED identifier reference set.
. July 2017 SNOMED CT International Edition: The antecedent SNOMED identifier reference set will be withdrawn and no longer be available in the release.

So the question for our AG is what are the best possible methods for removal of this refset?

I know that we've discussed in the past that completely removing items from the International Edition is problematic due to the removal of history, which contravenes our own standards to a certain extent.

However, this was a GA decision many years ago, and the fact that the refset is to be removed has been well publicised many times - it has also been stated that we will publish a static version in MLDS for future reference.

So in this particular instance, it's not a question of whether or not we should remove the refset, so much as what the cleanest method of doing so would be.

Some more technical information - we've been looking at the simpleMap files where the RT refset resides, and there are 432,470 records mapped to the RT Refset (900000000000498005).  This is expected, as this matches the total number of Concepts (minus one for the GMDN map concept to which RT wasn't mapped).

There are also, however, two additional records in the simpleMap file which reference the RT Refset, though not in the refsetID column, but instead in the referencedComponentID column:

82f3de7e-e15a-56c2-80a7-8ba824b9bd8e 20020131 1 900000000000207008 900000000000497000 900000000000498005 XUjjk
d9de6ab2-2243-56cb-9003-de0764f0236d 20020131 1 900000000000207008 900000000000498005 900000000000498005 R-5004A

Once of these is the mapping of the CTV3 Refset (900000000000497000), and the other appears to be a circular reference to itself!  So the question now is whether or not we remove these records from the simpleMap, in addition to the 432,470 records?

Also, do we surgically remove the supporting RT Refset metadata, such as the Concept (900000000000498005) and Descriptions, etc - thus creating a standalone static package for antecedent works?  Or do we simply remove the 432,470 simpleMap records into a separate package, and leave the supporting metadata in the International edition, meaning that the static package will forever remain dependent upon the International edition?

Please let me know your thoughts, as we need to decide this asap in order to implement the solution in the July 2017 release, and we therefore need to finalise the answer in the April meeting rather than starting the discussion...

Thanks very much!