Re: [Synaptic-devel] Newbie qstn: repository on a local hard disk?

[Panu Matilainen]

On Fri, 2004-07-30 at 03:52, Ari Torhamo wrote:
> WARNING, WARNING! NEWBIE STUFF AHEAD. This writing may be of interest
> for only those who want to know, what goes on in a newbies head when he
> is faced with something confusing. 
> 
> It's been a few days since I was told here how to create a repository in
> Synaptic of a local directory (thank you Michael Vogt and others). Now I
> have had time to try this, and I did succeed in creating one :-) It took
> me about two hours to accomplish - stop laughing... all of you... NOW! 

There's nothing to laugh about when a feature is so abysmally documented
as creation and use of local repositories is - the only docs I'm aware
of are the LWN-article Michael mentioned and various mailing list
archives :(

> 
> The thing that caused the delay, was that I had some wrong presumtions
> of the repository system. I thought that one has to know exactly how to
> name the local directories, and accordingly, what to write into the
> Distribution -field, when adding the local repository. I thought that
> there are some naming rules one has to go by, in order to get the
> installation program to work right and not maybe in the worst scenario
> to mess up the whole system. When I look at the http-repositories I have
> defined in (copied into) Synaptic, there seems to be a pattern in what
> they have inside the Distribution-field. It's mainly fedora/2/en/i386 or
> fedora/2. Michael Vogt used "connectiva" for this field in his
> instructions for me. I "knew", of course, that I can't use "connectiva".
> I'm not that naive, I thought :-) I was really puzzled by what I should
> put there. I surfed around the web trying to find this piece of
> information - what exactly to write into the distribution field when
> using Fedora and adding a local repository! Then, at some point, when
> going around clicking in the http-repositories, I started to think that
> what if there is no rule, maybe it's just a path! On the other hand this
> idea looked silly to me - why on earth would there be a special field
> called "Distribution", if it had no special purpose - other than just
> being a part of the directory path? Because I had nothing else to lose,
> but the system stability :) I decided to try. I just gave a random name
> for a repository folder and filled it in the Distribution-field. And
> heureka! There it was.

Yeah, for local repositories it's really just a path which has to be
split up in rather arbitrary way. 

> At the end the question remains, why has the repository path be divided
> in two. I did some experimenting and noticed that it makes no
> difference, how large chunk of the path you put into the URI-field and
> how large to the Distribution-field as long as you put the RPMS-part to
> the Section(s)-field. Maybe there are some historical, or other reasons
> for this - it would be interesting to know.

I'd say it largely originates from the way Debian distributions are laid
out and how APT was written with that in mind. The URI / Distribution
split makes sense for remote repositories but not really for local ones
and even less so for rpm-dir type. 

Local repositories could and should be made easier to use from Synaptic,
that and supporting installation of single arbitrary rpm's has been on
my todo-list for quite a while but haven't just had the time to do
anything but think about it :(

> 
> I know this story may seem silly and not so usefull from a developers
> point of view. But maybe it can serve as a small example of how one
> trivial thing can be a major obstacle for someone new to a program (and
> Linux).

Like Sebastian says your comments are not silly at all, it's a clear
usability problem with otherwise very useful feature.

        - Panu -