[josm-dev] validator question, multipolygons

[Dirk Stöcker]

On Fri, 4 Mar 2011, Frederik Ramm wrote:

> In my eyes the validator does not have a problem with one specific check; it 
> has an attitude problem. Until now I wasn't aware that it was *your* attitude 
> I was criticizing when I said so ;) but I think the validator is nannying 
> people too much, *especially* (and I checked that before writing it) since it 
> is enabled by default on a new install.

I asked some non-development users now and it seems there is an 
understanding problem between the way developers and users view the 
reports. I'm used to error, warning and info methods from a lot of 
development tools like compilers. But it seems most users don't understand 
a warning the same way. So probably we should find a better way to explain 
this situation.

> I don't even have to look past the warning dialog for my first complaint: 
> Even if the list contains only "warnings", the dialog title still reads: 
> "Data has errors." - That's what I mean by attitude problem; in my eyes it is 
> totally wrong to *ever* tell a mapper that his "data has errors". The 
> validator can at most point out potential problems - but "data has errors"? 
> As an expericed mapper I percieve that to be arrogance on JOSM's part, and as 
> a newbie mapper I would certainly not proceed with uploading.

I think after the next release (as it means translation issues) we should 
rephrase the wording and add some short descriptions texts, the validator 
is meant as an advince and it should not patronise.

When people misunderstand that we need to make it clearer.

> * "untagged way" (warning) - perfectly ok if such a way is a relation member. 
> You're not showing the warning if it is a multipolygon but there may be 
> others you don't know of.

I think this warning is nearly never a false positive. Thus warning status 
is valid.

> * "unknown relation type" (warning) - JOSM should never assume to be in 
> possession of a full list of allowed relation types!

Right, that it only knows certain types, but making it an Info-text makes 
it loosing its function. Here the "you should be sure if you know better" 
is required for users.

> * "unnamed ways" (warning) - I think it is perfectly normal to draw streets 
> from aerial imagery and have no name for them.

In most cases it is an mistake. Again this is a problem when the "warning" 
is misunderstood.

> * "illegal tag/value combinations" - someone seems to have had a field day 
> here. 90% of these deserve to be thrown out. Only recently it complained 
> about my "man_made=pipeline" - from reading the source I found out that it 
> was expecting an extra tag with details about the pipeline.

Yes. Here finetuning may be necessary. But this is a wide field and no 
generic solution, as each check has individual warning level.

> I think the main problem is that the validator is now by default enabled 
> before download - something you can switch off, of course, but to the new 
> mapper the "Your data has errors" message conveys: We don't want your data, 
> please stop what you're doing!

I agree, that can be a problem. So the solution must be to make clear what 
"error", "warning" and "other" of JOSM validator really mean. And we need 
to state that in few words, or people wont read it.

Also the online help for this needs extension probably.

> It is funny that both of us seem to have a desire to nanny JOSM users, just 
> whenever you're doing it I complain and vice versa.

Well, I think we agreed already in the past, that our opinions are equal 
in a lot of places, but we also have fields where we disagree a lot. Not 
that bad a situation at all, as usually it leads to results which are 
better than previous situation.

Ciao
-- 
http://www.dstoecker.eu/ (PGP key available)