[Monotone-devel] RE: Some thoughts on ids

[jcrisp]

Given your example, probably easier! Certainly the non-hash version below is 
easier to guess if you don't know... which is nigh on impossible with hashes. 
Maybe some kind of repository local mapping so that people can alias hashes to 
anything they like? Then you could alias a hash, and use forward and reverse 
navigation like [build01]+5 or something? I must admit my familiarity with 
monotone is low, so I don't know if you can search on date. I do think the idea 
of using hashes internally is a good one tho'
 
One key areas is labelling for builds, but if I've understood properly you can 
put a label on a whole hash tree representing a configuration, so that is OK.
 
BTW, the other things I *really* like about clearcase are traced managed builds 
(via the audited shell) and the ability to restrict the access to checking to 
the repository without a ticket in ClearQuest in the correct state.... ;-)
 
Cydergoth

        -----Original Message----- 
        From: graydon hoare [mailto:address@hidden 
        Sent: Tue 03/02/2004 16:47 
        To: address@hidden 
        Cc: address@hidden 
        Subject: Re: Some thoughts on ids
        
        

        address@hidden wrote:
        
        > Obviously there is a derivation difference in that the clearcase
         > IDs are generated by the database and are not tied to the contents
         > of the files. The reason why this may be important is that it may
         > be worth looking at the clearcase command set and seeing if there
         > are any things it does which are useful in terms of hiding the
         > underlying ID from the end user. The quick and dirty completion
         > stuff monotone already has looks dangerous to me....
        
        yeah, it might be nice to find a friendlier UI for referring to
        versions. I've considered maybe permitting a graph-path encoding with
        "anything that disambiguates" for each label (author, branchname, date,
        hash prefix, numbers assigned from 0..N based on lexicographic order of
        hashes), but I haven't given it a great deal of thought. I wonder if
        it's really easier to refer to friday:address@hidden:2a3 or just to
        get used to the SHA1 values in the few cases you actually need them.
        personally I find writing out arch's highly namespace-qualified
        identifiers cumbersome, but that's just me. I'm interested to hear
        preferences.
        
        so far my strategy is to try to reduce the number of times you have to
        refer to a version at all, and let the user read through "monotone log"
        if they need to pick one out in specific. but I'm open to suggestions
        (especially if patches accompany them).
        
        -graydon