Pre BK-7.0

  • Tags can’t start with a digit.

I think this is to make it easy to tell tags apart from revision numbers or from serial numbers.

  • Tags can’t start with the characters: @, =, -, +, or .

@ is due to ChangeSet revs and/or product revs.
= is due to serials
- is due to command line argument parsing (guess)
+ is to avoid confusing with + as tip.
. is to avoid ambiguity of A…​B; also want -r. to mean gfile

  • Tags can’t contain "..", ".,", ",.", or ",,"

This is all due to ranges.

  • Tags can’t contain the control characters: \001, |, \n, or \r

These just because they’re weird.

Problems

Dash is allowed in the middle of a tag, yet causes a parsing error in some cases, like bk edit -xA-B, but not bk cset -xA-B. Read slib.c:walkList; dash is used for a linear range. Deprecate that form of range? (which should be done anyway). Prioritize TAGS in the parsing?

Post BK-7.0

Same rules as above plus:

  • Tags can’t have a ? or *

? is when we don’t have a revision
* is due to regex/globs

  • Tags can’t contain ASCII characters lower than \040 or any of the characters in the following list: ``|?*\177~^:;/!\"$%&\\()[]{}` '

Note

All of these are trying to filter out the weird.