Is a vector map layer planned for openstreetmap.org?

I think there are a couple of reasons why a vector map of, what I assume you refer too, is the Standard or OpenStreetMap-Carto styled layer is not planned or realistic anywhere in the near future:

• As someone already pointed out, developer resources is one aspect. Knowing what I know from developing an OSM renderer from scratch for ArcGIS, it is an absolute huge undertaking to develop and maintain a style as complex as Carto / Standard, and the infrastructure to support it all. Converting the existing raster based style and infrastructure to a vector based one, would be a pretty massive task.

• There is a second, much underappreciated aspect though. Despite all the flashy promotion of vector based maps as being "light-weight" and "fast" compared to raster tile services, this is only so for very basic simple maps with just a few vector layers (maybe 5-10 max). As soon as you start converting a style as complex as Carto, which would probably require a couple of hundreds of vector layers, things start to look entirely different...

• Also, Carto currently displays buildings from a scale of about 1:50k. That would mean literally (tens of) thousands of vector building objects needing to be transferred and rendered on mobile devices with big screens if a 1:1 vector style conversion would need to be implemented with the same level of detail as Carto at each Z zoomlevel...

Honestly, have you ever seen a style even nearly as complex as Carto being served as pure vector tiles?

I haven't, I think that says enough. I think current mobile hardware simply is not yet up to the task to deal with such complex vector services. Maybe in another 10 years...

Just testing of some vector services on my Samsung Tab S (first version), which is pretty high end as regards mobile with an octo-core processor, shows map rendering in the mobile browser to be very sluggish compared to opening the same vector service in a browser on my Core i5 desktop. And that is with a very basic style. Opening the Google Maps App, which does offer acceptable performance, and zooming all the way in, I only get to see any buildings when there are no more than maybe 50-500 buildings max in the viewport...

Also, in this respect, I recently read an article where the author concluded that with the apparent switch to vector based services, a style like Google's was considerably simplified compared to the styling before the switch. Less objects were shown at each zoom level. So you can have vector on your mobile device now, but at the huge cost of great simplification of the styling and the number of features shown at each zoom level.

Is that the price you want to pay for having vector?

Just to chime in here. I've been working on porting the openstreetmap-carto style to vector tiles. But to render it server side with mapnik, producing images at the end, not to use client side rendering. It took a bit of work, but it can be done. I am not sure if it's sensible to have that rendered client side.

Well, oh my oh my, what was I wrong!!

I am taking back my words about not being able to publish a 350+ layer Vector Tiles service...

I knew ESRI was working on Vector Tiles, but hadn't yet found the time to explore the new addition in Pro 1.4. I am not used to ESRI getting it right the first time, it usually takes a second round.

But I just published an Amsterdam OpenStreetMap extract to a Vector Tile service created with ArcGIS Pro, with all freaking 350+ thematic layers!

There are some caveats though: - (Open cross) hatch polygon symbols are not yet supported in Vector Tiles, they end up being solid fills, which covers up stuff below it, which it really shouldn't. - Also hatched line symbols fail, and draw as a simple line feature - Random stipple pattern symbols seem replaced with equally spaced ones, meaning lesser cartography. - Symbols that use a reference scale for their size (e.g. 1:10k), and should be scaled in size / made bigger or smaller at other scales, don't. This doesn't work with Vector Tiles and the current client rendering. - Labelling is not as good as ESRI's Maplex label engine, there is quite a lot of labels missing.!But all in all, I am surprised to see this output so easily! I even published in a custom UTM zone 31N projection, instead of default Web Mercator.

The attached image shows both the Vector Tile (Right) and a PDF of the same data as comparison (Left). Notice the solid fill instead of cross hatch near Amsterdam Central Station, and a lot of missing labels.

Lastly, my first generation Tab S was able to handle opening this smaller extract as well.

No, there is no such planning.