W3C

– DRAFT –
FHIR RDF

07 April 2022

Attendees

Present
Dagmar, David Booth, EricP, Gaurav Vaidya
Regrets
-
Chair
David Booth
Scribe
dbooth

Meeting minutes

Concept IRIs

dbooth: I think eric's position is that we should allow creativity in the design of concept IRIs, which would require that things like slashes get passed through without percent encoding.

eric: yes.

dbooth: I have sympathy for supporting creativity, but also worry that the designers of these terminologies will probably not be thinking about how their codes will be turned into IRIs.

eric: We could define two different algorithms, with a flag for each stem IRI, that says which algorithm to use.

dbooth: Confirmed that a hash cannot appear in a fragid.

eric: We could more fully parameterize the algorithm depending on what part of the IRI it falls into: path, search parameter, fragid, etc.

eric: SNOMED has a hierarchy, eg a procedure or an abstract value or reference range -- they all have their own places in the SNOMED hierarchy. If Kent Spackman did it again, he could take that into account.

gaurav: If codesystem (or a new property that goes with stem IRI) tells you how to validate the code, then that will tell you how we need to encode it?
… E.g., if SNOMED says that slash is allowed in codes, then we can pass it through without encoding it.
… For most wiki codes, they give a regex and a template.

gaurav: For FHIR codesystem, you can define a filter to only included certain codes.
… They also allow regex to filter them.
… And you can have multiple filters, and 8-9 operators and they don't expect implementers to implement them all.

dbooth: Exactly which chars are we concerned about? Only slash, and always encode hashes? Or are there other other chars we might want to pass through?

eric: hash, question mark, ampersand, semicolon, equals, slash.

eric: For SNOMED you'd want to encode everything, because if somebody used a postcoordinated code w comments in it, that should be percent encoded.

dbooth: Is there other data that would help us decide? Look at what chars are used in codes?

gaurav: UCUM has slashes, square brackets, asterisk, percent (for percentages), curly brackets.

dbooth: That's an example of slashes clearly not being used for hierarchy -- they're used for division.

eric: There's also two schools of though about where to put units -- into the quoted part of the RDF term, vs into the datatype.

UCUM examples:

h
U
ug
[IU]
mg/m2
ml/h
kcal
cal/mL
mL/d
mL/h
cal/[foz_us]
[foz_us]
{scoop}
{score}

dbooth: I think we need more evidence about what chars are used in codes.

eric: Could have a conservative algo and a liberal algo.

gaurav: I like that idea. But an example had old codes that were no longer resolvable, eg LOINC code: LA6721-0
… Could say that this algo only applies if the code matches ##### (all digits)

ACTION: Gaurav to find codes in more terminologies to see what chars they use.

dagmar: In our work we should always end a stem IRI with slash or hash.

ADJOURNED

Summary of action items

  1. Gaurav to find codes in more terminologies to see what chars they use.
Minutes manually created (not a transcript), formatted by scribe.perl version 185 (Thu Dec 2 18:51:55 2021 UTC).

Diagnostics

No scribenick or scribe found. Guessed: dbooth

Maybe present: dbooth, eric, gaurav