[Osmf-talk] Tiles

[Brian M. Sperlongano]

On Thu, Jun 1, 2023 at 10:09 PM Mikel Maron <mikel.maron at gmail.com> wrote:

> The upside to vector tiles is flexibility in rendering for various
> mapper use cases, aiding data creation. One idea I heard recently was to
> show a map where older features fade from the map. Pretty cool idea. Of
> course ultimately  depends on what tags and metadata are included within
> the tiles (you can do everything) but something like last edit time seems
> sensible property potentially.
>

I'm a big proponent of vector tiles.  After all, I helped kick-start a
vector-tile-based community map style[1].

I've also observed that the idea that the OSMF needs to bankroll vector
tile technology to jump-start style development hasn't aged well. When the
idea was first proposed, rendering vector tiles took a long time and
serious hardware. We needed the OSMF to host vector tiles because it was
unreasonable for a casual developer to pay for a high-end server to do the
rendering or rent a cloud server to do it.  However, technology has
changed, and it was because individual software developers solved problems
they had and not because the Foundation kick-started anything.  It's now
possible to render a planet vector tile server for $1-2 of cloud compute
time and about $10 a month to store and host it online.  This is how I host
a vector tile server[2] for my map style for the cost of a couple of cups
of coffee a month.  Of course, if it gets too popular, bandwidth costs
money, and I'd be forced to cut it off.  However, these same economic
considerations apply to anyone wanting to dabble in vector-tile-style
development and they'd be able to spin up their own.

Another observation about vector tiles is that if the tile schema is not
tailored to the style, you waste A LOT of bandwidth serving up tile data
that isn't displayed to the users.  Sometimes that's okay, such as serving
tiles with multiple languages so the user can switch the display in their
browser. But if the tiles have certain data at zoom X, but the style
doesn't begin displaying that data until zoom Y, or not at all, that's a
waste of bandwidth, not to mention unnecessary slower load times in the
client. So, the idea of a centrally-hosted "general purpose" tile set
becomes questionable for non-trivial styles.

I acknowledge that there are technology gaps to address.  For example,
minutely vector-tile updates are not there yet (full planet renders in
about 30 minutes have been demonstrated), and I understand that there are
folks working on it. If the Foundation focused on closing technology gaps
like that, I'd be more interested, though I'm not sure that would be the
best use of limited funds.

In summary, technological advancements have made vector tile rendering and
hosting more affordable for individual developers, reducing the need for
the OSMF to support style development financially. While there are still
technical challenges, it is debatable whether it is the OSMF's
responsibility to address them, given the many other competing priorities.

[1] https://zelonewolf.github.io/openstreetmap-americana
[2] https://6ug7hetxl9.execute-api.us-east-2.amazonaws.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/osmf-talk/attachments/20230601/12b154ea/attachment.htm>