[OSM-talk] Proliferation of path vs. footway

[Frederik Ramm]

Hi,

David Earl wrote:
> Up to now, we could get away with changing existing tags, but as people 
> start to use OSM for real world tasks and base software on it that is 
> outside the OSM community, like other file formats, we really have to be 
> more controlled about upward compatibility and support. People won't use 
> OSM if we keep changing it in unexpected ways under their feet.

My view is that we should aim to make the best data set we can. We may 
have to compromise in many ways, e.g. we cannot make things to difficult 
or we lose mappers; we cannot be more precise than GPS allows; and so 
on. But I am very reluctant to let our work - at this early stage at 
least - be influenced by what we think the "users" might want. Because 
all this "backward compatibility" for the sake of some guy somewhere who 
has written a program that he's unwilling to adapt really really hurts.

Next thing you tell me is that in the future we won't be able to make an 
API change without providing lots of unnecessary and slow code for 
"backwards compatibility".

I think that OSM should concentrate on their core competency and that it 
collecting data. If we ever find we have to change everything and do 
things in another way because that gives us better data, then by all 
means let's do that and not be encumbered by some users down the line. 
If they want our cool & free data they will have to take it in the form 
it comes, or employ/pay someone to make them something that does not 
change. But I am absolutely unwilling to take our flexibility away for 
the sake of "customer service" - because I think that what we first and 
foremost deliver to our "customers" is good map data, and we must not 
choose any inferior way of collecting that data just because we believe 
that the current "customers" like that better.

> We may realise we made a mistake doing something on way and not another, 
> but we have to take into account the impact and cost of changes against 
> the perceived problem something causes.

We should always take into account what changes for us and our mappers, 
but anything on the "user side" I think the users can take care of 
themselves. (In the end they will benefit from better data as well, it 
would only be short-term interests that would make anyone demand that we 
remain backwards compatible.)

> For example, changing highway=gate to barrier=gate. That allowed for a 
> consistent way of presenting barriers, but at the expense of anyone 
> relying on gates to block routing through them for example not working 
> (and if they weren't aware of the change - why should they be? - theior 
> programs stopped being effective). This was a largely cosmetic change, a 
> change for tidiness sake, not because it was necessary.

I'm all for making such changes, in this case I think it was mainly an 
improvement for mappers (because USERS don't see the tags usually). If a 
router cannot be bothered to change then the software is probably not 
suitable for OSM. OSM is not an ISO standard ;-)

(But of course we could think about having some kind of announce list 
which those only using our data could subscirbe to to be alerted to 
changes they perhaps need to follow.)

> Changing the common tags like footway/path or the main highway 
> designations would be a disaster for these reasons.

I don't currently support any idea of changing these and honestly I 
doubt I ever will. But if there were a proposal for changing them, I 
maintain that it must be evaluated without regard to downstream users - 
just for us "makers", and if it is deemed to bring improvements for us 
but alienate the users, then by all means implement it. (Someone will 
undoubtedly be quick to offer some kind of backwards-compatible 
synthesized planet dump for those that cannot adapt. Nothing that we 
need to worry about.)

> Consumers (people and software) want to have confidence in what we 
> provide. 

If you want a rigid structure and reliability then get your data from 
someone who's been doing it for 20 years, not 5. Because if we attempt 
to "freeze" now we'll be miserable for the rest of our lives.

Bye
Frederik