From freedesktop-dbus  Thu Jun 15 19:21:00 2006
From: Thiago Macieira <thiago.macieira () trolltech ! com>
Date: Thu, 15 Jun 2006 19:21:00 +0000
To: freedesktop-dbus
Subject: Re: Escaping for paths
Message-Id: <200606152121.00492.thiago.macieira () trolltech ! com>
X-MARC-Message: https://marc.info/?l=freedesktop-dbus&m=117323265230781
MIME-Version: 1
Content-Type: multipart/mixed; boundary="--nextPart2520434.9zQXfb8pgh"

--nextPart2520434.9zQXfb8pgh
Content-Type: text/plain;
  charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

Ross Burton wrote:
>Hi,
>
>Several applications are writing methods to escape strings so that they
>can be used in object paths (specifically Network Manager does this, and
>eds-dbus has taken code from NM too).  EDS is using it so that it can
>return an object path of the
>form /org/gnome/evolution/addressbook/[uri], where obviously the URI
>needs to be escaped.
>
>I think it would be useful for DBus to contain a function that escaped
>an arbitrary string so that it could be used in an object path to avoid
>this needless wheel re-inventing.  Anyone else agree?

I discussed this very issue with another developer yesterday. The issue in=
=20
question was bookmarks instead of addressbooks, but it's exactly the same=20
use-case.

My conclusion is that we should not provide such a generic method. I=20
believe it's dangerous. I'd much rather see the applications internally=20
changed to avoid this kind of problem.

In this example, what I would instead propose is that the object=20
at /org/gnome/evolution/addressbook provide a method called "FindURI" or=20
something similar, which would take the URI and return an object path to=20
the internal object. This would have the added benefit of on-demand=20
loading, since there's a call before the actual object access.

I also don't believe this kind of functionality belongs in the core=20
library. Validation? Yes. Escaping? I don't think so.

The reason why I think it's dangerous is that two different URIs or=20
pathnames could escape to the same object path. For example, imagine I=20
had addressbooks:
  /home/thiago/address_book-vcf
  /home/thiago/address_book.vcf
(that's a Greek alpha and a Greek beta)

Both would escape to /home/thiago/address_book_vcf or=20
_home_thiago_addres_book_vcf.=20

The only solution would be to select one of the characters we have to be=20
an escape character, like % in URIs. The above example would=20
be /home/thiago/address_5fbook_2dvcf=20
and /home/thiago/address_5fbook_2evcf, or=20
_2fhome_2fthiago_2faddress_5fbook_2dvcf and=20
_2fhome_2fthiago_2faddress_5fbook_2evcf. If people push arbitrary data=20
into this function, it would have the potential to increase the length=20
threefold.

Or to completely escape everything (yielding something ugly like:=20
_2f686f6d652f74686961676f2f61646472657373626f6f6b2e7663660a)

=2D-=20
Thiago Jos=C3=A9 Macieira - thiago.macieira AT trolltech.com
Trolltech ASA - Sandakerveien 116, NO-0402 Oslo, Norway

--nextPart2520434.9zQXfb8pgh
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)

iD8DBQBEkbMcM/XwBW70U1gRAmm7AKCH31rGZysAiDYuv1cbaNwpI5CQsQCguzZV
eP98Zp2qUTvis0hfCrxa7HA=
=lPlf
-----END PGP SIGNATURE-----

--nextPart2520434.9zQXfb8pgh--