Re: major libgweather Locations updates

[Dan Winship <danw gnome org>]

Djihed Afifi wrote:
> First, are you absolutely sure this is stable now? I won't touch this
> until it is.

No, that's why I suggested people should wait a week or so before making
major effort at translating it. Because as people report problems, I'm
going to try to make improvements.

> I am not sure what
> methodology was used to prune or consider cities.

For the most part, it is the same as before; we list cities where there
are weather stations. The only difference is that now:

    - We also try to include major cities that don't have their own
      weather stations

    - We try to have each location be a city name, rather than an
      airport name or something else

    - If there is a location in a small town, and there's a medium-sized
      city nearby, we try to name the location after the medium-sized
      city instead of the small town. (This depends on us having good
      data about the relative sizes of cities, so it works better in
      some countries than others.)

Fundamentally, things are still driven by where there are weather
stations reporting.

> I had a quick look at Arab cities in general and Algerian cities in particular.

So as a first pass, you sould just be comparing to GNOME 2.22, rather
than to the theoretical perfect list of Algerian cities. You'll find
that the list hasn't changed all that much (and hopefully where it has
changed, it's better now. Eg, Algiers has been added,
"Tamanrasset/Aguenna" (a combined city and airport name) has now become
"Tamanrasset" (just the city name), etc.

> Algeria also has a very odd population density. 80% of people live in
> 15% of the land. But more than half the cities in the file are Saharan
> cities.

Yeah, so this is because that's where the weather stations are. From
googling/wikipedia-ing, it looks like a lot of them are associated with
oil fields. If they really are basically useless, we can remove them
from the list. (Vincent and I were having this debate on IRC the other
day, about a handful of weather stations in very very tiny towns (eg,
~100 people) in France. On the one hand, it seems better to keep
them--maybe the computers at the oil field run GNOME? :) But on the
other hand, more locations means more work for translators, so I can
understand the desire to keep the list small...)

> The city choice was not made on population either.

For countries where we have population data, all cities with population
greater than 100,000 are included. However, we don't have population
data for very many countries. We can fix this manually by adding cities
to libgweather/data/major-cities.txt.

> In brief, some Algerian choosing cities, and looking under Algeria
> will be wondering how exactly were the cities chosen. I suspect the
> same will be true for many people around the world.

Yes, but that was true in GNOME 2.22 too. This is a first step toward
improving it.

As for your question of what to do with suggested improvements, please
don't make changes directly to Locations.xml.in at this time; put the
fixes in a bug report instead.

-- Dan