Register name/identifier

[arnau]

First, when we name a Register, we identify it so there is an underlying issue with this overlap.

The current naming convention says that a register for a thing, say Foo, should be the name of the thing, so in this case it would be foo. This assumes Foo is managed by a single custodian across the UK.

If each part of the UK were managing Foo on their own, we would use foo-eng, foo-wls, foo-nir and foo-sct.

The current convention has a few issues or undefined behaviour:

• What happens if Foo is managed by one custodian for more than one part and for other custodians for the rest of parts? Being more specific:
  • 4 parts, 1 custodian. Covered, it's foo.
  • 1 part, 1 custodian. Covered, it's foo-eng, foo-wls, foo-nir and foo-sct.
  • 3 parts, 1 custodian + 1 part, 1 custodian. Undefined.
  • 2 parts, 1 custodian + 2 parts, 1 custodian. Undefined.
• What happens if the custodian owns less than a part of the UK?
• What happens if the scope changes over time? E.g. we start with foo because it applies to the UK but at some point one part starts to handle it on its own. E.g. you start with foo and end up with foo + foo-sct? What does it mean for foo to be now a 3/4 UK rather than all of it?
• Extending on the above, what happens if sct and nir start handling Foo on their own? is foo still a reasonable name for the register?

/cc @nacnudus I tried to summarise here the naming issue. Let me know if I missed anything fundamental.

[Jamie-Atkinson]

principal-local-authority is not local-authority-wls as it is not a register of local authorities in Wales, there are hundreds of them. it is a register of principal local authorities which are different

[arnau]

Removed the bit about principal-local-authority 👍

[nacnudus]

We also don't define eng, sct, nir, or wls anywhere (no UK register).

[Jamie-Atkinson]

the way i have been thinking about it is majority rule. If a register contains at least 3/4 then it is foo and the descriptions reflects that it covers only 3/4 the 4th is then added as foo-nir etc.

if it is a 2:2 split then both would recieve the foo-sct-eng

1/4 foo-sct

2/4 foo-sct-eng

3/4 foo

4/4 foo

(and we always put the -eng-nir in alphabetical order)

[arnau]

@Jamie-Atkinson in your approach what happens if the scope of foo changes over time?

[Jamie-Atkinson]

@arnau with the way I think about this it would no longer be the same register. It ties both the thing (e.g. biscuits) and its restrictions (e.g. only in England) together to make the register.

under this model, if we changed the scope of the register it is no longer the same register. We would retire the old register and create a new one with the new restrictions.

If we remove the attachment of the restriction from the register definition then this model breaks. Unless we used meaningless ID's for the registers (as you suggested in our f2f) then I don't see how we could name for this.

If we made the register: schools and stated in the description it is only for England, then ignoring the issue of what do we call schools in Scotland, how can anyone build a service on it. If the contents of the register might change from day to day how do I build a service with the data? If I want to capture which school you go to in England, and then my dataset evolves to include Scotland how can I plan for that? Even worse if we at a later date stop including England in the dataset.

[arnau]

Ok, makes sense. The concern then is more about name clashes/availability over time and consequently name inconsistencies due some retired register owning a name that would fit better with a new register. This on its own makes me think that a register identifier shouldn't be meaningful.

The logical next question is how to keep a reasonable readable domain names. I need to think a bit on this.