What type of matches do we really need?

[wilaw]

The current draft proposes match types of

Prefix - match from start
Exact - exact match
Suffix - match from end
Matches - mid-content match.

Comments have been made that MOQT specifies only prefix matching for NAMEPACES and NAMES and therefore that this is all we need on the auth side too.

However a counter argument is that we need more flexibility when specifying constraints than we do when matching namespaces.

My recommendation would be to

1. Keep PREFIX and EXACT - these are well aligned with MOQT semantics
2. Drop SUFFIX and MATCH as I can't see a use case for these

[alficles]

One typical use for "SUFFIX" is when extensions or names are used to provide information about the resource. For example, in HTTP, you can often say something like:

{
    /catu/ 312: { /path/ 3: { /prefix/ 1: "/path/to/folder/" } },
    /or/ TBD_OR: [
        { /catu/ 312: { /path/ 3: { /suffix/ 2: "/manifest.m3u8" } } },
        { /catu/ 312: { /path/ 3: { /suffix/ 2: ".ts" } } }
   ]
}

In HTTP, the prefix usually represents some kind of "category" for the content and the suffix tends to have information about the "kind" of the content, with stuff like extension or file name. None of these are hard-and-fast rules, but because people usually structure things that way make both prefix and suffix useful.

MATCH does mostly exist for completeness, though. It would be pretty easy to drop. (Though it's also easy to perform, cause it's doable with a simple scan.)

[wilaw]

In HTTP, the prefix usually represents some kind of "category" for the content and the suffix tends to have information about the "kind" of the content, with stuff like extension or file name. None of these are hard-and-fast rules, but because people usually structure things that way make both prefix and suffix useful.

In MOQT however, there are no trailing extensions or the equivalent of file extensions. All tuple field matches are left-matches for both namespace and name.

A quesiton I have is - are contiguous tuple fields matched, or partial fields?

For example, is "example" a prefix match for "example.com/cusomerid/appID", assuming the field boundary is represented by a "/" ?

[suhasHere]

For token matching, I feel have parts of tuples matched is useful .. something like
/abc/def//123