[Talk-us] Highway Tagging Consensus to Improve OSM (and address some of 41 latitude's concerns)

[Nathan Edgars II]

On Mon, Oct 18, 2010 at 8:50 PM, Ian Dees <ian.dees at gmail.com> wrote:
> On Mon, Oct 18, 2010 at 7:43 PM, Nathan Edgars II <neroute2 at gmail.com>
> wrote:
>>
>> On Mon, Oct 18, 2010 at 8:38 PM, Ian Dees <ian.dees at gmail.com> wrote:
>> > On Mon, Oct 18, 2010 at 7:32 PM, Nathan Edgars II <neroute2 at gmail.com>
>> > wrote:
>> >>
>> >> On Mon, Oct 18, 2010 at 8:15 PM, Ian Dees <ian.dees at gmail.com> wrote:
>> >> > Ways and relations are equally easy to "break"
>> >>
>> >> Nope. They're not, because relation membership is not a tag on the
>> >> way.
>> >
>> > Yes, they are, because a relation membership and new way (with new tags
>> > and
>> > node refs) are both completely separate new entities that need to be
>> > uploaded. If an editor can reliably do it with a new way then it can
>> > also
>> > reliably do it with a new relation member.
>>
>> When a way is downloaded from the API, its tags are all downloaded,
>> but a separate API query has to be sent to get any referring
>> relations.
>
> Presumably if you are splitting a way you will have already downloaded at
> least the one relation way member that you want to split.

But you haven't necessarily downloaded the relation itself. On the
other hand, it's impossible to download a way without its tags.
>> >>
>> >> A way is always uploaded with its tags,
>> >
>> > Yes, but the new (split) way is not necessarily always uploaded.
>>
>> Relations are uploaded after new ways to avoid precondition failures.
>> Thus it's much more likely that a connection will fail between the way
>> and the relation than between two ways. I've seen Potlatch take
>> several tries to upload a large relation, which compounds the possible
>> problem. It's also obvious from the map when a portion of a way is
>> accidentally deleted, but even if relations were used to render
>> shields an error would not always be obvious.
>
> If the renderer uses route relations to render roads, why would errors be
> any less obvious than when ways were used?
>
If the new (split) way is not uploaded, it shows up as a gap in the
road. That's going to be really obvious.
>>
>> I have seen these problems in my work on repairing relations. Do you
>> have any such experience?
>
> I have been offering suggestions to solve the route problems in the original
> thread. Do you have any such suggestions?

I have no problem with the solution of using relations, but ref tags
on ways should remain as a backup.