i'd reserve that for designated public parking, and not knowing the area, i have no idea if that is the case, or if this is used just as ad hoc parking by employees.

there are designated carparks at the south, but it is hard for me to guess how far the general public has access with a motorised vehicle.

i'd also guess from the name as a camp, it is less intended for public camping, but for the yoofs who may not yet be of driving age (though i had noted for y'awl, driving ages seemed to be going down and drinking ages up, the reverse of the situation over here, but that was decades ago).

in any case, in a changeset comment yesterday i read confirmed that at least one pedestrian routing engine fails to handle areas. then again, i see parking aisles being drawn through carparks.

honestly, i don't particularly care. i performed my vandalism, i can see the result, so i'll go bother some pokemon parks springing up in preparation for next month's yearly ingress database import.

if you would, take a look into
way/654529301
it has only a postcode, no other address info, and as opposed to the angle of the rest of the buildings, this one seems completely aligned n-s-e-w, and indeed, the nodes of the middle horizontal line have in common
Location: 38.6638070,
as seen in the ESRI imagery, here are two buildings that do not resemble this polygon, so i'm wondering if this was how the data was delivered by microsoft, or if somehow it was twisted to perfection.

curiouser and curiouser.

sure felt like i was deleting a hundred ways in the same place, all sharing the same points. i *would* have lost count, and did take a break for a partial upload, moved on, then returned. it was that bad, after i did count 19 ways at one white-roofed building. overall, 144 ways slain, but only 90 nodes. i know osm should strive for efficiency, but *this* is ridiculous.

the addresses at that one spot were from several different streets. essentially, anything i deleted here is a bug not from the microsoft side, but in some cases, the addresses should have been left as nodes (where two or so apply to one building), but something is seriously pear-shaped in several other locations, like a church with quite a lot of address ways overlaid on themselves.

there's getting to be not much left to judge positional accuracy, as most of these were not errors in the outline data, but errors in assigning addresses, which really needs to be found and fixed before any further production upload.

i need a break, resolving spam notes or something, for a bit...

i suspect i overlooked one or two, but these are the outlines where i would draw two buildings, which have been assigned more than one address, resulting in multiple overlapping ways sharing the same nodes.

one might expect these to have two addresses, and in fact a few did. but many had three or more -- up to ten or so, i actually lost count while deleting.

obviously trying to merge addresses with outlines here is destined to failure.

it;s hard to say how many locations were afflicted, without looking at the data -- 33 addresses, but only 60 nodes, so maximum 15 locations, probably 10 or 12 or so.

i guess next (or when i wake up) i will see how many of what i would trace as a single-building outline, has multiple addresses, purging them...

by the way, i'm analysing your data and commenting on it as i purge it bit by bit.

it's probably easiest if you look at my most recent changesets, or else look at my changesets i've commented or those in the us.

links to help you find this drivel would be
@freebeer/history
(valid for the moment - stop reading as soon as i mention Pokémon)

https://resultmaps.neis-one.org/osm-discussion-comments?uid=429761
(does not include the changesets above where i have not commented yet, one so far)

i am finding more problems i did not expect, in particular the overlapping outlines for multiple buildings with differing addresses. that looks like something important to be addressed, although the worse problem is where the Microsoft algorithm is interpreting and combining multiple separate building outlines into one large one.

i trust you are not surprised to learn i find several buildings demolished since the data was collected - i'll make a pass for those, although it will be limited by excluding anything that has collapsed in the past year or two since the mapbox aerials were flown.

i hope what i am doing here will help you to find problems in your data preparation, that should be addressed before further uploads.

while the roof of one of the buildings is intact in the Bing z20 overflight, it is seen to have collapsed in later Mapbox. DG Standard does not show any difference...

This building outline consisted of three overlapping ways, all sharing the same points, but with differing address data, so there's a problem assigning addresses to polygons that incorrectly span multiple buildings.

Address 4239
way/654529367/history

Address 4233
way/654529374/history

Address 4235
way/654529370/history

I am pretty sure i will be seeing the same thing in the overlapping 2-in-1 and 3-in-1 and 4-in-1 outlines which i have not tried to delete yet.

am i having fun yet?

answers on a postcard, please

one outline deleted with four distinct building outlines within; one more to be analysed.

urgh, was hoping to use digitalglobe standard as possibly newer than mapbox as determined elsewhere, but it is just too grotty for even me.

here 16 2-buildings-in-one outlines - but worse, many more overlapping outlines as i described with my earlier 3-building delete, probably comparable in number to this. aieee.

also saw one or two outlines i should have included in my 3-in-one deletion, this time using somewhat more recent mapbox imagery, which also shows several buildings that have (unsurprisingly) been razed since the Microsoft datagrab, which may well correspond with their overflight which i guess to be about five years old by now, and the mapbox imagery may be getting on to two years by now. just wild-arse guessing.

five standalone outlines in the Microsoft building dataset for this neighbourhood included what i believe to be three separate buildings each.

in addition, two more outlines encompassing three buildings were not simple outlines, but look to be multiple ones overlapping or something, but i don't see the highlighted co-positioned nodes i expect, so i'll have to examine those in detail in a later changeset.

no, my suggestion is *not* to try correctting this data. it's too far gone and would be a waste of time.

pretend the data you uploaded no longer exists. as soon as i get enough free memory, i'll go through things deleting them and collecting statistics. i'll do the revert for you, slowly.

i'll let you know when it is safe to upload here again; if you get things working and want to upload in a different non-overlapping neighbourhood, go ahead.

what does this #hashtag #salad have to do with this edit? this is nowhere near indonesia and says nothing about this edit.

unfortunately, unrepresentative changesets like this, or the one to have placed a giant mosque atop the White House and throughout Ireland, do nothing to help the already-tattered HOT reputation amongst the OSM community, when these mapping efforts are supposed to be confined to helping countries in need, not damaging places around the world.

these sorts of edits, the exception and not the rule, but too often to overlook, do a great disservice to the vast majority of HOT mappers and volunteers who *do* take mapping and quality more seriously, and i don't know whether they are indicative of a failure of the review process, or a failure in training, or both.

if i may suggest, hold off on a revert. the data is not hurting anyone and cannot be seen. if anything, it's an ideal testbed for someone wanting to analyse the quality of the microsoft buildings, as the above comments from me might cause someone (anyone?) to have a look-see.

actually, i'm thinking of going through and manually deleting things in several passes to get an idea of the problems in the data, and at the end, after counting what remains, revert the lot.

specifically, first i'd delete multiple buildings counted as one, probably starting with five-as-one down to two-as-one.

then i'd nuke others that fail some degree of accuracy. all in different changesets, to be able to count how many victims fell by the wayside, and what could be salvaged.

i am afraid i cannot help with JOSM questions, as i have never used it and most likely never will. pity, for then i would know if one can apply a slight rotation transform to the buildings and salvage them, something i cannot do in potlatch.

the purpose of my testing would be to see if it is worth machine-adding this data and most importantly, *cleaning afterwards* for the most egregious of errors, or if that janitorial duty would be time best spent in manual tracing. other reviews i have seen have spoken of the quality not being suitable for a blind import - particularly if city or country building data can be made available as in a few other parts of the us.

so, i hope i have your blessing to experiment here, and i thank you for giving me the opportunity to do something other than provide free customer support for the facebook/instagram map errors being dumped into the osm support channels.

again, if someone doesn't chime in with JOSM tips, try asking somewhere else, and i apologise for not being useful there.

thanks

Hallo, lyft mapping team.

may i ask, what tool(s) are you using to identify and locate broken relations?

and how is the migration to editor JOSM in place of iD going? i see iD is still being used by the last edits of your team to have caught my eye.

yhanks

peeling meself once more from me deathbed to shuffle over to the privvy, it occurs that me alzheimers is taking its toll, and i should have added more golden rules.

do not demand nor expect perfection from anyone but yourself. strive yourself for perfection. when you reach it, decide it isn't good enough and start over. let those who are without sin revert the first stone. keep mapping-related discussions in a public place where other sane and stable members of the community can contribute and point out my obvious bull-puckey. and never ever *ever* take anything i write or say seriously. that way be dragons.

keep an eye on developments in other parts of the world if you can understand the language; this changeset discussion just popped up today and is relevant here:
changeset/65566428

your secret santa^H^H^H^H^H undisclosed member of the Data Werking Grope,
freebear

NURSE! THE MEDS!!! BRING A MOP, I'VE DROOLED INTO ME KEZBOARD AGAIN

this is bing z20, which i know to be pretty good. it also matches the new mapbox.

yeah, both the archived bing (clarity) and esri have extreme offsets. esri matches dg premium; dg std also has an offset which one can expect, given the lack of care that goes into those mass coverage aerials from an angle in what must be a hilly region here, and minimal attempts to align things.

i'd bet your WAAS-adjusted GPS traces will come close to the carefully-prepared z20Bing and mapbox.

esri clarity (bing archive) is usually pretty good when everything else is dire, but it fails here, but luckily the bing z20 uses a much better terrain model.

if it exists in reality, there is no reason not to have it on the map, and here it looks like a valuable addition when mapped and tagged correctly.

secondly, in germany, there is at least one firm which is paying to map private driveways as they are important to its business (foot access); same looks to be true elsewhere for amazon logistics and other services for their delivery drivers who presumably would in fact be able to use these private drives.

before any revert of this import is carried out, i'd ask you to take a look at the uploaded data and see if it meets the standards you expect.

way/654530292
is one of the outlines i randomly picked to view, and it looks a bit at a funny angle compared with what i expect.

your data is not rendered because there is nothing to identify it save for the address info in the wrong format, but if you want to see everything, try
https://nrenner.github.io/achavi/?changeset=65416347
where you can zoom in for a rough preview of how your data will appear without further processing, which in some areas (without comparing against imagery) is, well, um ...

i will refrain from digging in with both hands into the can of worms already open in the mailing list and devouring them with relish. or horseradish. kremžská hořčice, aww, now i'm hungry again. and i had hoped to give up eating until Lent.

in the upper right of the added buildings i see that four separate buildings were combined as one. top row, three as one or two as one. actually, all over i see that, so i'd give this import a 2 of 5, as there are some decent reasonably well-oriented outlines that match and could be clicked through, but i'd personally discard many or most of them as not meeting my personal standards of craft mapping.

just thought i'd bring this up before you get too far. for reference, my background imagery is the Bing z20 found in metro areas like this and which appears to be quite well aligned with a similar LIDAR or whatever process used to distort the features, viewed with the `dim' checkbox in my potlatch 2 selected.

Personally, i'd let humans map most of the duds, but then in this area or in East St Louis i skipped a lot as without JOSM my outlines could not meet my standards.

five separate buildings with one postal address. i wonder which goes where.

and it's a pity, for there are so many unmapped simple rectangular outlines that were a delight for me to trace even before the Mapbox imagery was released. ah, the good old days when i actually contributed...

do you mean to revert the entirety of this edit, so you can start anew, re-creating the objects but this time with the correct tagging values? presumably you can just pull the same objects from the MS database with no real extra work.

i can do that for you, no problem

or do you mean using josm to convert the tagging errors into something correct?

i can't help there...

Just to add an irrelevant data point, I would not put too much faith into Auntie Beeb - pull up a bollard and let me tell you a tale.

I was listening recently to BBC radio when they had a short news report on Karl-Marx-Stadt and their newsreader came up with a complete howler. how embarrassing.

i suspect that the newsreader was promptly marched into one of the subterranean gulags of the BBC and for the next 55 minutes underwent the traditional re-education course and grueling torture, for the next news bulletin did not repeat the same error.

The same risk applies to the case here -- i bet the same american-illiterates-mangling that Karl-Marx-Stadt suffered, will be applied to Kiev, by all but the 1% who actually pay any attention to the world. I mean, if you can even distract their attention away from their phones, and ask them to speak the written word, i bet the majority of them are going to mistake it for the australian chanteuse Kylie Minogue and autocomplete it to that or at least pronounce it in a similar manner, which bears no relation to the pronunciation you likely expect.

I would expect the majority of them to be unable to speak the name of Karl-Marx-Stadt without mangling it exactly as that BBC newsreader's first failed attempt.

By the way, the BBC is the first hit i get for `Chicken Kiev' as well.

obviously the solution would be for americans to grow a familiarity with the rest of the world, but realistically there is no longer any chance of that, and things including spelling have to be dumbed down to their level, as there is too much ambiguity between written and spoken english.

there is a place for insults and pseudo-superiority in osm, but my comments are not that place.

Carry On Arguing.

well, the rendering works. now there seems to be a dancing headless chicken (put yo' wings up in the air) for which i accuse the poor resolution of the archival bing clarity and it has nothing to do with the established fact that i am a lousy mapper in the first place. Mapbox, your failure to blanket the entire world with low-flying drones has irreparably ruined my reputation. Now see if you can sleep nights with that burden on your consciences.

routing, i dunno; i hope we as a civilisation have not sunk so collectively low that we rely on assisted navigation to get around recreational areas and wonder how elders like myself even managed to survive with our LED-free whale-oil lamps and interactive maps carved into our cave walls.

nonetheless, if only as an experiment before unleashing these spindly-leggy can-can chickens upon the world at large, because i see people continuing to draw footpaths across pedestrian plazas to make every possible route option visible, i should probably investigate the legitimacy of leaving your labours intact and slathering pirouetting poultry atop, seeing i wanted to leave an easy escape when my play inevitably turns pear-shaped.

if it's legit, that chicken can be de-boned of the skeleton you so painstakingly traced to replace the tangled spiderweb of yore.

or better yet, revert my changeset which i have assisted social-media addicts to suscirbe to with appropriate hashtags (changeset comments are so yesterday and should be abolished)