Format of URLs - The given URLs are ’overlayed’ according to their
priority, and they get a name (to ease updating only parts).
Such an extended URL has the form
where URL is a standard URL known by subversion -- something like
http://...., svn://... or svn+ssh://....
The arguments before the URL are optional and can be in any order; the
URL must be last.
or, using abbreviations,
Please mind that the full syntax is in lower case, whereas the
abbreviations are capitalized!
Internally the : is looked for, and if the part before this character
is a known keyword, it is used.
As soon as we find an unknown keyword we treat it as an URL, ie. stop
The priority is in reverse numeric order - the lower the number, the
higher the priority. (See url__current_has_precedence() )
Why a priority?
When we have to overlay several URLs, we have to know which URL takes
precedence - in case the same entry is in more than one. (Which is not
Why a name?
We need a name, so that the user can say ’<b>commit all outstanding
changes to the repository at URL x</b>’, without having to remember the
full URL. After all, this URL should already be known, as there’s a
list of URLs to update from.
You should only use alphanumeric characters and the underscore here;
or, in other words, \w or [a-zA-Z0-9_]. (Whitespace, comma and
semicolon get used as separators.)
What can I do with the target revision?
Using the target revision you can tell fsvs that it should use the
given revision number as destination revision - so update would go
there, but not further. Please note that the given revision number
overrides the -r parameter; this sets the destination for all URLs.
The default target is HEAD.
In subversion you can enter URL@revision - this syntax may be
implemented in fsvs too. (But it has the problem, that as soon as
you have a @ in the URL, you must give the target revision
There’s an additional internal number - why that?
This internal number is not for use by the user. It is just used to
have an unique identifier for an URL, without using the full string.
On my system the package names are on average 12.3 characters long
(1024 packages with 12629 bytes, including newline):
COLUMNS=200 dpkg-query -l | cut -c5- | cut -f1 -d’ ’ | wc
So if we store an id of the url instead of the name, we have approx. 4
bytes per entry (length of strings of numbers from 1 to 1024). Whereas
we’d now use 12.3 characters, that’s a difference of 8.3 per entry.
Multiplied with 150 000 entries we get about 1MB difference in filesize
of the dir-file. Not really small ...
Currently we use about 92 bytes per entry. So we’d (unnecessarily)
increase the size by about 10%.
That’s why there’s an url_t::internal_number.