Having more than three phases (was RFC: Phase UI (revset, phase command and others))

[Jason Harris]

On Jan 9, 2012, at 1:07 AM, Steven Brown wrote:

> On 9 January 2012 02:56, Jason Harris <jason at jasonfharris.com> wrote:
>> 
>> On Jan 8, 2012, at 2:51 PM, Pierre-Yves David wrote:
>> 
>>> On Sun, Jan 08, 2012 at 11:59:43AM +0100, Laurens Holst wrote:
>>>> Op 7-1-2012 18:05, David Pierre-Yves schreef:
>>>>> 
>>>>> Secret changeset are not indiscoverable secret in many ways. For
>>>>> example you can pull them anyway if you explicitly  ask the server
>>>>> for them. The secret phase is more a safety to prevent unexpected
>>>>> pushing/pull some changeset, relying on them to keep stuff
>>>>> undisclosed is wrong assumption and the name is probably bad then.
>>>> 
>>>> But the cryptographic nature of changeset IDs prevents this from
>>>> exposing information about the existence of these changeset IDs.
>>> 
>>> I do not understand what you mean there
>>> 
>>>> Of course it is true that it’s not really great that if you somehow
>>>> retrieved the changeset ID, you are able to retrieve the whole
>>>> changeset. So if you’d ask me this shouldn’t be possible either
>>>> (what’s the use case for allowing it?).
>>>> 
>>>> Why is this a wrong assumption, why wouldn’t we guarantee that
>>>> secret changesets are actually that, secret? I think that would be a
>>>> great and very useful property.
>>> 
>>> The current mercurial code base work that way. The implementation of "secret
>>> changeset" so far does no do anything else that pruning them from discovery
>>> mechanisme.
>> 
>> So again I think the name "secret" is then bad. Laurens has taken the command
>> at it's face value, (as I would) and then is rightly complaining that something secret
>> is not actually "secret". Better names might be:
>> 
>> private / draft / public
>> 
>> private / draft / complete
>> 
>> unshared / draft / complete
>> 
>> etc. But unless secret is actually, well secret, then it's likely not such a good name.
>> 
>> (BTW thanks for going through all this work! )
> 
> Secret is an intuitive name for me. One of the things you can do with
> a secret is to tell it to someone.
> 
> I find that private and unshared are less accurate. They both imply
> that no-one else can see it.

But I have seen tables like:

On Jan 3, 2012, at 2:24 AM, Olle wrote:

> I like the idea of private and draft, perhaps the word to be used can
> come from the opposite of draft?
> 
>          immutable shared
>  complete:     X        X
>  draft:               X
>  private:
> 
> or maybe:
> 
>          immutable shared
>  absolute:     X        X
>  draft:               X
>  private:

And I had assumed this was a correct reflection of what was going on. And David
has just confirmed that secret things are not really secret... So it seems to me
secret is a bad word choice here. I guess the question really comes down to
(i) are these changesets meant to be inaccessible in all ways to other people
(secret) or (ii) are they meant to by default not be seen by others but can be
seen / accessed if chosen to. In which case these are unshared / private
changesets...

Secret things I would think would be generally much more highly "classified"
than private or unshared... But I guess David might have to comment here.

Cheers,
  Jas