I don't think we want to get into the
registrar business. A vendor can figure out how to create a unique
function name if they want to create such an extension, whether 123.DAYS,
or 123_DAYS, or whatever. If we want to recommend a mechanism, then
that's fine. But creating a registry is overkill. As you said,
the number of applications is relatively small.
I look to XPath as a good implementation
example here. They allow custom functions, and allow them to be namespaced.
But over the years I have not heard of problems caused by vendors
colliding in names.
-Rob
"David A. Wheeler" <>
wrote on 07/22/2006 02:01:15 PM:
> Different applications have different functions, and we want
> applications to be free to create new functions. Yet the semantics
> they create may be incompatible with each other (for example,
> Lotus' 1-2-3 DAYS is incompatible with OOo's DAYS function).
>
> We need a way to identify functions created by a given application,
> one that won't interfere with others. This is one of the
> requirements in our charter.
>
> I propose a very simple approach: Already, some applications are
> using "." as a name space separator for function names
> (see SQL.stuff, ERROR.stuff, etc.). So, let's simply assign
> name prefixes for each application we know of, and a way
> to get your own name prefix formally assigned.
>
> That way, you can have "GNUMERIC.XOR(...)" as a function.
> A given application probably would not show the prefix for
> itself (so for GNUMERIC, this would probably just show as XOR).
> Proposed predefined prefixes: EXCEL, LOTUS, GNUMERIC,
> OOO, KSPREAD, WIKICALC, QUATTRO. They could nest, so
> EXCEL.SQL.QUERY() could be Excel's SQL.QUERY() function.
>
> I don't think we need to use XML namespaces for this.
> The number of applications and independent libraries is
> relatively small, and we could probably have OASIS act as
> the registrar for names.
>
