Vector tiles on OSMF hardware

Hello all,

as a follow up to the [vector tiles discussed previously]), they are now running on OSMF hardware. The usage policy is not yet final, but you can use it now. The tilejson is at https://vector.openstreetmap.org/shortbread_v1/tilejson.json and the tiles are at https://vector.openstreetmap.org/shortbread_v1/{z}/{x}/{y}.mvt.

This is a final production location, so you can build stuff based off of it. I’m sure there are still some bugs, but this soft launch should reveal them.

The demo page also is using the new OSMF-hosted tiles. It feels fairly snappy to me, as I would expect when most tiles are pre-generated. There’s some work still to be done, and I have a temporary issue tracker acting as a to-do list. The main areas outstanding are

  • monitoring,
  • cache headers,
  • low zoom rerendering, and
  • publishing the usage policy.
30 Likes

Here’s the previous thread: https://community.openstreetmap.org/t/minutely-updated-vector-tiles-demo

Paul, what is the plan going forward with respect to multi-lingual (name) tags support? This is the thing that most people wanted vector tiles for (misguided or not), so I would think that at least some more languages should be supported than just German and English.

PS: there was a longish discussion of the ins and outs of this in the previous thread starting in June, however there was no clear conclusion.

8 Likes

That isn’t quite clear, Is the statement

“there is currently no published usage policy”

or

“there is a usage policy, but it might change”

If it is the later, having access to it would be helpful.

How often does this layer get updated? Is it taking minutely diffs.and if so how long are cached tiles expected to hang around before replacement?

1 Like

That was all answered in the original thread linked above, but anyway see Minutely updated vector tiles demo

1 Like

That was all answered in the original thread linked above, but anyway see Minutely updated vector tiles demo

it says minutely updated, but the tiles I just got from here
https://pnorman.github.io/tilekiln-shortbread-demo/#17.88/41.860264/12.494244

are older, maybe there is some caching happening somewhere?
See in OpenStreetMap:
https://www.openstreetmap.org/#map=19/41.860138/12.494310
The way in question was changed 5 days ago: https://www.openstreetmap.org/way/44327374

In general it is an advantage to actually read the thread Vector tiles on OSMF hardware that you are posting in.

Which part of the current thread do you think answers that question?

The original posting in this thread by Paul announcing the production site, making questions about Pauls private test site absurdly superfluous.

5 Likes

Post 1 give a couple of addresses for the tiles for us to test in our own stuff [1] and then says:

Which is a link to: https://pnorman.github.io/tilekiln-shortbread-demo/ exactly the same site as:

So those should be the tiles we’ve been told about in the first post???


  1. FWIW it works in QGIS, but I’m having either style translation issues or a serious case of user error when it comes to the icons for the colorful style. ↩︎

3 Likes

Here’s what I see:

It seems like the tiles reflect your removal of the oneway=yes tag but not your earlier deletion of the one-way roundabout approach and outlet.

In case it helps, my OSM–OHM mashup demo comes with the maplibre-gl-inspect plugin. Click the button in the upper-right corner to see an X-ray view of the tile contents.

Where should we report styling problems? (This thread?)

1 Like

If they’re definitely bugs then I think the correct place is the Versatiles github repository. (Or at least I hope so as I made one for it rendering some piers the same colour as the underlying water which seems to be an error rather than a choice.)

If it’s more to do with choices and omissions made by the style then I don’t know. That might be a thing for the Spirit style rather than Versatiles Styles? What is actually going to be used when we get to the stage where there are vector stiles on OpenStreetMap.org?

This post seems to lay out the various places to report the various issues, but I don’t know if some or all parts of it have been overridden in the subsequent ~100 posts.

On a slight tangent: is anyone else seeing fine lines at the tile boundaries for the land/kind=forest polygons? e.g. here I don’t know if this is a style issue or an issue with the way Maplibre is behaving in my browser(s):
image

I don’t know if my eyes are playing tricks on me but I think I’m also getting blurrier looking icons than they appear to be in @Minh_Nguyen’s post am I missing a pale halo that’s meant to give an unsharp mask-like effect or is this a jpeg compression thing? (or am I crazy?)

image

(I know these aren’t technically tile issues. Please direct me to the right thread, if there is one.)

It isn’t - Versatiles is a separate project and not what Paul is working on here.

So style issues shouldn’t be reported anywhere?

I asked the same question many months ago and could not receive an answer. I wanted to report some service ways missing, but couldn’t. Posting again here, perhaps I’ll be luckier this time.

They told me it’s not shortbread nor Versatiles, so I’m a bit lost here.

In this case, what is missing is the rendering of ways with highway=service + service=driveway.

In the tiles linked at the top of the post I am seeing examples of:

streets 
    bridge: false
    kind: service
    oneway: false
    oneway_reverse: false
    rail: false
    service: driveway
    tunnel: false

So that’s not a shortbread limitation or a limitation of whatever software is creating the vector.openstreetmap.org/shortbread_v1 tiles. It just happens to not render on the demo page that uses those, which at present is using Versatiles colorful.

What I’m struggling to find in the 60 page megathread is an indication of what the initial style will be when (if?) this (eventually) gets incorporated into the main website and where is best to put things on the wish list.

Post 92 on that thread says:

So if nothing has changed it will eventually be the Street Spirit style[1], preceded by some unknown style using the style independent shortbread tiles.

As far as I can tell there is no interim style (or fork thereof) named as the one that will be used initially so there is nothing to put requests to that is OSM affiliated.


  1. The only preview I can see of this is here. ↩︎

1 Like

I am seeing some very odd (caching?) things happening at the moment.

Last night I did some editing that added a beach (w1333595340) and a few other things including water bodies.

This was partially visible about four hours later at zoom 15.62 and I grabbed a screenshot:

and also partly visible at zoom 13.35:
image

Today however I can only see the beach and canal at zoom 13.x and they have disappeared again at z14+:
image

Another browser’s incognito mode, QGIS and a separate Android tablet show pre-beach-edit tiles at zoom 14.

Additional Firefox weirdness

An additional layer of confusing for me is that the pnorman.github.io/tilekiln-shortbread-demo/ demo and the https://1ec5.github.io/tilekiln-shortbread-demo/ when viewed in Firefox seem to be showing me different cached versions of the same tile zoom 13 column 2374 row 3517 one without and one with the beach (respectively) even though they are both ostensibly the same file. I have deleted cookies and site data for both and then done a hard refresh with the cache disabled in the Network inspector and this persists. Neither site is showing the beach in a Chrome (or Microsoft Edge) incognito window so I think this bit is probably Firefox being weird and insisting on showing and old thing when it has been instructed to get it fresh.

Is there any reason why I would be getting tiles with old data having previously received tiles with newer data?