[Telepathy] straw man location API

mikhail.zabaluev at nokia.com mikhail.zabaluev at nokia.com
Wed Jan 30 04:38:17 PST 2008


Hi, 

>-----Original Message-----
>From: telepathy-bounces at lists.freedesktop.org 
>[mailto:telepathy-bounces at lists.freedesktop.org] On Behalf Of 
>ext Simon McVittie
>Sent: Tuesday, January 29, 2008 4:09 PM
>To: telepathy at lists.freedesktop.org
>Subject: Re: [Telepathy] straw man location API
>
>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>On Tue, 29 Jan 2008 at 15:42:37 +0200, 
>mikhail.zabaluev at nokia.com wrote:
>> GeoClue have made an unwieldy API for it with rigid flat 
>structures and
>> bit flags, Win32 style :-/
>
>There's nothing wrong with bitfield flags - Telepathy has lots of them.
>They're an efficient way to transfer, well, sets of flags.

When flags designate which other fields of the structure have valid data, that's what I refer to as Win32-ness. It's a lot like doing variant types by hand, and it doesn't really win much compared to general D-Bus overhead. A matter of taste, probably.

>I don't like some of their API (human-readable strings over D-Bus?
>erm... are the backends expected to be localized?... and, as you say,
>their civic location structs are rather rigid) but it's better
>than inventing our own *again*. Have you brought up your concerns with
>the GeoClue people?

I was going to, and I probably will soon.
As for civic location strings, let's suppose for sanity sake that only the data source does any possible localization.
It's probably the only way with most location databases of the day.

>I still maintain that unless GeoClue is so terrible that we 
>consider it a
>complete dead-end (and IMO that's not true), the way forward 
>is for people
>interested in Telepathy+geoloc to treat them as "the Telepathy 
>of geolocation"
>and use their formats by reference; and if their formats are 
>crap, help them
>to fix that before they freeze D-Bus API.

I've briefly reviewed their new design proposal, and it looked like they design for a single daemon providing the device's location data. I should recheck, but I believe there are no data types defined for locations to pass around.

There is also some minor ugliness in the D-Bus APIs, like passing three separate arrays of the same size where an array of structures would seem logical.

-- 
  Mikhail


More information about the Telepathy mailing list