maelito

Was it tested on a road in a car ?

Main developer of cartes.app here. We've internationalised the app. Expect some traces of French but we're working on it :)

But yes, the development force remains mostly French for now. Waiting for more European contributions...

https://translate.codeberg.org

You're right. Last holidays, I went to beautiful place in Spain. A way covered in Ferns was marked as an OSM path. I guess it was, but did not get cut this year or the later. So we tried and failed 30 m after.

It's important to keep some mysteries despite precise maps.

That's why I'm at the same time fascinated by 3D "splats", but also worried to be spoiled.

They're great for places we already know (inspect your roof's shape for instance), but they spoil discovery.

It's basically Comaps but not open-source, right ?

In the app I'm developing, cartes.app, we acknowledged this lack of pictures.

We show local Wikimedia common pictures, osm tag pictures, og:image pictures of the website if any, Wikipedia article infobox pictures, Panoramax for street-view, and last but not least, any picture dropped by an ATproto place review.

Lots remain to be done, especially building the latter community of reviews.

One of these viewpoints maybe ?

https://cartes.app/?cat=point-de-vue#13.66/62.56098/7.70188

Thanks a lot !

Could you try https://cartes.app and share your search terms and expectations ? We're using Photon, an OpenSearch search engine. We've build a dictionary of place categories to help in-browser with FuseJS.

Also thinking about using embeddings to complement this.

Hi ! Yes, was freezed by gedankenstuecke's comment. But yes you're right, I'll write an update.

I'm not talking about stars, we don't care about stars for search engines.

What's important is people talking about "has air conditioner" in a hotel review or "has vegan" in a restaurant. OSM has none of this data.

Awesome. We have a dictionary of words to OpenStreetMap tags here : https://codeberg.org/cartes/web/src/branch/master/components...

Do you think your work could help us let users type "pancake" and get "crêpe" without writing an explicit "pancake = crêpe" dictionary entry ?

In practice : if I understand well, your lib would first need to download 5 Mb, once and for all, and would then be used as we use Fuse.js right now ?

How well does it handle languages other than English ?

Could it be "trained" on the OpenStreetMap tag wiki ?

Thanks a lot for your work.

OSM is mostly a db with an old website for contributors, osm.org.

It's as fresh as OSM databse : it depends.

For good (server) search, one needs many layers (Photon, Pelias with OpenAdresses, Overpass, in-house pmtiles, etc.) using many DBs, each needing server ressources or expensive paid APIs.

It's obvioulsy expensive in terms of ops + dev, but also just to host.

It can't scale with only 0,0001 % of users donating to the app.

Fortunately, NLNet's there to fund work, but it's still nonethless only a tenth of what would be needed.

Plus map applications and general search engines don't talk to each other... I don't know why, but it is so. Maybe because all the well-known search engines are closed-source ?

It doesn't take so much to enable a good server search on top of OSM + openaddresses.

Local search will always be slow and bad.

But server search doesn't need that much. It's just that OS initiatives are severely understaffed. OS apps that have a Photon instance are already rare to find. Let's not talk about having an Overpass instance...

What is very hard to reproduce is Google's place review data.

It's golden to enable good search.

Who also build Tilemaker.

We at cartes.app have discussed with some people Comaps. We've made links from cartes.app to Comaps, e.g. for navigation, a feature that we don't have.

We've not seen anything from their side. We don't really understand why they wouldn't want a proper Web version.

It's even worse in the non-open source world : Kagi's maps are bad. Qwant dropped their cool map project, too expensive. Ecosia has none, redirects to Google, its main competitor.

They could "just" clone cartes.app and run `bun i && bun start` and `ansible-playbook` to launch their instance of cartes.app, but we can't reach them. Or fund the project.

What's important though is that we collaborate on the tools we all use. OSM, MapLibre, Photon, Pelias, Protomaps, Motis, Brouter, etc. are all tools that we somehow share, each end-user app having their own combination of them.

We also thought like you do when outside of this world. But now, its a human problem. How to reach people and how to gain the time that it would take to collaborate ? We don't have the keys.

Last I checked, Osmand uses an unconventional map rendered. I would want to contribute to it, it's lagging on so many devices, I can barely use it. Comaps is perfectly smooth.

I'm not saying ego problems don't exist of course. But taking Osmand and making a simple general public app out of this is a huge effort (human and tech), probably more than building a side app.

That makes sens. Both features combined would be cool. In any case, a cache invalidation feature should be always visible to the user, so that he knows that clicking this button will switch to online mode.

Yes, it's a similar project. We share the same stack.

Checkout the French equivalent : https://carto.tchoo.net. Looks more complete.

Past similar HN submission got no attention, whereas the UK's top page. Interesting !

https://news.ycombinator.com/item?id=45249351

Did you explore the solution of just caching the pmtiles requests responses as well, instead of IndexedDb ?

No. You cannot even click on a bakery to open its page. It's a bad Web client for the general public, and the vector style is hidden all below. Osm.org is the reference website, the one where all data is always up to date and above all, where you edit the map. But it's not a good OSM client.

Funny : there's another Ben that did his "Ben maps". https://benmaps.fr

Awesome, thanks for the links.

Yes it's planned. Should work in theory, but no one tried before.

Maps are so often done as native apps that to my knowledge no one tried to just use the PWA capabilities to cache tiles on the Web. Of course what's hard is cache invalidation. Does the user want to update the tiles ? Never ? Daily, weekly ? Only some regions ? Or manually ?

Here's an issue about that : https://codeberg.org/cartes/web/issues/1078

It's in French, unfortunately Codeberg has no auto-translate capabilities yet.

Organic Maps and its fork Comaps still lack a Web client.

We're this on https://cartes.app, trying to push the Web further (even on mobile devices) so that you don't even need an app for most use cases.

We the FOSS world just have 100 to 10000 less budget to implement this feature. It's a hard problem, and you can't ignore that lots of other features are missing. E.g. street view or place comments.

I'm working on https://cartes.app and we're well aware that search is not on par, far from it. But we have hundreds of other features and bugs to fix. https://codeberg.org/cartes/web

Did you try it on an old device ? The experience on a less than 5 years old device should be quite fluid.

The mobile sheet maybe ? This one is hard : there is no good Web mobile sheet available yet...

Shameless plug : I'm working on https://cartes.app, a Web OpenStreetMap app.

Far less difficult as coding for an old OS obviously, but still a challenge !

The Web has plenty of potential, but constraints too, mostly because of dominant actors, such as Apple that hid the PWA install button... Or Firefox not having any install banner, whereas chrome does.

Can't post it yet as a proper subject, it can't handle top page load.