[josm-dev] validator question, multipolygons

[Dirk Stöcker]

On Sat, 5 Mar 2011, Frederik Ramm wrote:

>>  If I judge this issue based on the ticket reports we get, than we have
>>  only minor problems with this. And half of the reports ask to add
>>  additional checks and not to remove some.
>
> That's because you have created a perfect user nannying environment and 
> people react to that. It's an interesting challenge. Once you see that the 
> validator will complain about certain things in a multipolygon relation but 
> not about others, you immediately think: "Oh well it should actually also 
> complain about this, and that. And if someone does that together with that we 
> should warn them too. And..."

This is actually irrelevant. The Trac bug report system only catches these 
people who are willing to report anyway. So the absolute count does not 
say a lot (only compared to other reports), but the type of reports we get 
is a useful measure.

> I'm 100% sure that the people who ask for more checks are *not* those who 
> say: This error always happens to me and I'd really be happy if the editor 
> could tell me.

Well - Actually here you are wrong. One example is a person who wants the 
coastline checks moved from OTHER to WARN level, as it still happens often 
that coasts are destroyed.

> These are the people who say "I really think the others should be told 
> to map like I do."

Very likely also this kind exists, but we aren't forced to agree to their 
request.

>>  So a note to these of you trying to convince me that we have a major
>>  problem with validator: This opinion does not match the statistical data
>>  that we have. Especially as validator had >80% installation count even
>>  before it moved into core.
>
> I'm not sure these statistics can be applied the way you apply them. How did 
> you measure installation count? How many of those were auto-installed through

I cannot measure installation count.

> the Windows installer which AFAIK bundled the validator with JOSM? At that

Yes. Sure. And exactly this invalidates the opinion that moving validator 
into core changed the situation. Effectively it has been a de-facto core 
component for a long time now.

> time, was the validator already set to validate before uploading by default?

This is default as long as I'm involved in JOSM development.

> How has the number and extend of checks be changed between then and now? Etc.

The number of tests has been increased. The number of tests the users sees 
in default settings has been decreased a lot.

Whether you and others like it or not - The opinion expressed on 
mailinglists like this (and talk-de and others) is not equal to the 
real-world. Thus I do not base my judgment on these discussions, but I try 
to find better data. And if that data shows me a totally different view, 
then I'm very sceptical.

Software users aren't that dumb as most want to tell me. They are 
usually much more flexible that most of you seem to assume.

P.S. That doesn't mean there shouldn't be improvements in this area.

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