From: Oren Ben-K. <or...@ri...> - 2002-02-10 06:39:12
|
Clark C . Evans wrote: > 1. I don't particularly like !\com.clarkevans.timesheet, That would be !/com.clarkevans.timesheet > If there isn't a colon and a slash and more than one period, > "com.clarkevans.timesheet", then its normalized form should > be prepended with "yaml:any/" to make: > > com.clarkevans.ts -> yaml:any/com.clarkevans.ts > > I know this prevents us from using "!date.value" as > a short hand for "yaml:any:/org.yaml.date.value", but > I don't see this as a big use case, where the above > one is a big use case. I don't know... I rather like the '/', it makes it clear the name is "absolute" as opposed to being "relative to yaml.org". As for its usefulness, we don't really know right now. We may want, for example, to have !seq.fix and !seq.var to distinguish between "array" and "vector"; or !int.32 and !int.big, etc. I'd rather not close the door on this. > 2. I was wondering if we could switch the role of the > backtick(`) and the pipe (|) characters? So that the > pipe is used to split the format, where the backtick > is used for the short-hand mechanism. The pipe is > a very nice separator and is not very magical; where > most people expect magic when they see a back-tick. Hmmm. So: !int`dec !/com.clarkevans.|ts !|ts Would become: !int|dec !/com.clarkevans.`ts !`ts I agree | is nicer than ` by far. How about we just avoid ` altogether by using ^ instead? !int|dec !/com.clarkevans.^ts !^ts This has double "intuitive" mnemonic. First, '^' stands for "start of match" in regular expressions, which is suitable for a prefix-related indicator. It also points up/back to the ancestor node defining the prefix. I'll go for | for format and ^ for prefix. Thoughts? > 3. I was wondering how Brian's perl works with this new spec. I did change some things (like allowing trailing spaces and adding new escape sequences). Other than that it is what we agreed upon. Of course if we change the characters he'll have to change his code and test data accordingly... If we do want to change the characters I suggest we do so ASAP. I really want to be able to point people at a spec I can trust not to change next week. I suspect Brian shares the sentiment :-) I suggest we settle this and announce an "implementation freeze" like we tried to do a few months back. We should also publish a link to it in SML-DEV and maybe XML-DEV and ask for comments (I wonder what the XML-DEV folk would make of it :-), with the understanding that we won't be changing it for a while - say, three months? then, when we have some code and some feedback, we can consider further changes. It would do us good :-) Have fun, Oren Ben-Kiki |