Vector tiles on OSMF hardware

InsertUser · 2024-11-12T12:57:21.397Z

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?

SimonPoole · 2024-11-12T13:35:45.010Z

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

dieterdreist · 2024-11-12T16:57:36.000Z

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

SimonPoole · 2024-11-12T17:20:23.796Z

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

InsertUser · 2024-11-12T17:40:50.159Z

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

SimonPoole · 2024-11-12T18:59:14.377Z

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

InsertUser · 2024-11-12T19:06:08.329Z

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

pnorman:

The demo page also is using the new OSMF-hosted tiles

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

dieterdreist:

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

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

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. ↩︎

Minh_Nguyen · 2024-11-12T21:00:38.067Z

Here’s what I see:

Viale Guglielmo Massaia as a two-way street, beginning at a one-way outlet from a roundabout.1920×1073 57.3 KB

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.

An X-ray view of the tile contents around Viale Guglielmo Massaia, detailing the feature properties on that street.1920×1073 98.3 KB

arctic-rocinante · 2024-11-12T21:06:30.288Z

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

InsertUser · 2024-11-13T14:46:52.629Z

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):

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?)

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

Richard · 2024-11-13T14:50:33.978Z

InsertUser:

If they’re definitely bugs then I think the correct place is the Versatiles github repository.

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

InsertUser · 2024-11-13T15:09:36.686Z

So style issues shouldn’t be reported anywhere?

matheusgomesms · 2024-11-13T15:35:54.269Z

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.

InsertUser · 2024-11-13T17:12:43.450Z

matheusgomesms:

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:

Minutely updated vector tiles demo

Assuming the second part is picked up by the OSMF it will be Shortbread initially followed up Street Spirit.

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.

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

InsertUser · 2024-11-14T15:41:09.473Z

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:

image686×572 26.5 KB

and also partly visible at zoom 13.35:

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

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?

pnorman · 2024-11-15T21:58:07.286Z

SimonPoole:

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.

I have opened an issue in Shortbread for this.

SimonPoole:

That isn’t quite clear, Is the statement

There is no approved usage policy. You can still use the service right now.

InsertUser:

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?

It is consuming minutely diffs. We have not yet tweaked caching settings so don’t know how long they will be held in the CDN.

Various style questions

The style the demo page is using is Versatiles Colorful. Style bugs should be reported there. Do not report vector tile issues there. If the tiles produced do not follow the Shortbread Vector Tile Schema 1.0 this should be reported with the definitions. Change suggestions for the Shortbread specification should be reported to its github page.

Errors serving (e.g. 5xx tiles, etc) can be reported here. They might be an issue in Tilekiln, the backend server setup, or the Fastly setup.

InsertUser:

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

I’m not sure what OSM affiliated means in this context. None of the styles of the featured layers on osm.org are controlled by the OSMF or the general OSM community, they are all the responsibility of their maintainers.

InsertUser:

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

One of the servers has stalled in updates and I haven’t yet investigated. The monitoring (not yet enabled) would have shown this.

PeachyOne · 2024-11-15T22:32:48.736Z

The style may not be controlled by OSM or the OSMF, but Carto Is run on OSMF hardware and is propagated largely because it is on said hardware.
The vector tiles would be much the same and its pretty normal for it to be given a similar degree of scrutiny.

SimonPoole · 2024-11-16T10:11:22.766Z

Actually you can use any shortbread compatible mapbox-gl style you want.

While we are kind of assuming that the OSMF might provide a selection of such styles together with the vector tile service, it isn’t a given (and they are not doing it at this point in time).

frodrigo · 2024-11-17T14:09:10.619Z

pnorman:

~~https://vector.openstreetmap.org/shortbread_v1/{z}/{x}/{y}.mvt~~

Just a remark. Please, try to avoid promoting the direct tiles URL, but in place promote the tilejson URL. We have now a new chance to avoid the attribution drama. As the tilejson include the attribution, and show it by default on map (like in MapLibre GL).

InsertUser · 2024-11-17T16:22:50.490Z

Not everything that support vector tiles supports tilejson.

for example:

Making people parse JSON in their heads isn’t the most user friendly strategy.