What makes for osm.wiki/Good_changeset_comments

What makes for osm.wiki/Good_changeset_comments

▪︎what makes for osm.wiki/Good_changeset_comments
▪︎osm.wiki/Changeset#Geographical_size_of_changesets

New hotness 👍🙂.

👍🙂

“Thank you for a very respectful answer.”

Likewise; I forgot to mention, in my previous reply, that I'm thankful for your being receptive. Sadly, many aren't.

“I have never been to Jersey myself, but I have many connections there. One of my connections lives in the area I modified, and my tagging is based on what he have told me.”

That's vastly better than most non-local mappers, who often guess & assume, or otherwise treat Jersey as part of the UK (which it isn't).

“It was very difficult to align the imagery, and my additions are not based on the same alignment (yikes).”

I can imagine. It's a mess. Nothing's aligned (with anything else).

Until more GNSS traces are published, the least-bad option is probably to align with existing data, as you say. Then, in future, someone brave can mass-realign whole areas without having to do too much checking of source:* values.

Unfortunately, the early mappers don't have the (relevant) technical background to appreciate the problems or considerations. Hence silliness like tagging for a particular renderer (for example, quite a few nodes have addr:housename merely to make that (descriptive) text appear in the tiles, instead of being properly tagged (and (gasp) using description=* for descriptions), recognising that rendering is secondary to data-integrity, and that the database itself is the map, not any particular rendering; but, alas; been fixing plenty of that, too, as I find & survey it).

Compare Guernsey (to the north-west); they've at least done things more in the right order, and have plenty of traces everywhere.

“Houses are added straight from the Bing alignment as I couldn't find better alignment, and I don't trust the current roads to be aligned.”

Many roads aren't well-aligned. Often things are offset by several metres. Bing isn't terrible, but not is it close. It's better than it used to be (I recall reading about 100-metre offsets(!)).

At present, other than collecting more traces, there's no good option. So, best-effort, and tagging additions clearly for the benefit of whoever tries to organise it properly.

But, at least misalignment is a known problem. I've had arguments with over-zealots folks who ranked an outdated almanac higher than my very-recent survey. Frustrating. Particularly when such folks value uniformity (of tagging) over accuracy. Sigh.

“When adding roads or messing with their geometry I first aligned the imagery with the existing roads. This is to move existing roads as little as possible, and still have the added geometry make sense.”

Yup, probably the best than can be done, at present. Likely what I'd do. Often what I actually do (because, e.g., placing a small feature according to GNSS would make it appear somewhere obviously bad, like the middle of a road, or the other side of a wall).

A lot of these sorts of problems are only apparent due to having GNSS for real-time comparison.

Makes for headaches, at times. Imagine trying to survey a row of houses (for addr:housen*, building:* & roof:*) when the alignment to GNSS is off by 5+ metres, and so when standing in front of one house, the map shows me as between two houses (or when between two, showing me as in front of one). Trying to keep track of which is which, while also looking for the values for tags, is a pain. Sometimes I've given up (and occasionally left a frustrated note).

I do wonder, sometimes, if misalignment is why Jersey gets little attention from foreign armchair mappers; maybe they simply give up.

Hence my suggestions about using source:geometry. If I know which imagery was used, then that makes surveying a little easier, when there's significant misalignment.

“I'm hoping to send someone out with a GPS later this week, and see if I can get enough traces to actually align the imagery correctly.”

Ah, that would be good. I should upload the several dozen I've captured since my last upload.

Would you have them tag their uploads so that they're easy to find (amid the noise of all the others, especially New Jersey). For example, with ‘Jersey’. See the traces under my profile, for examples.

“I agree that access=destination is probably better.”

👍

I use OsmAnd for navigation, and it will avoid (by default) ways marked as private, but allow those with most other values for access. Amusing, sometimes, when it tries to send me down a way which is clearly private, but hasn't been tagged as such (nor with the gates at each end) 😄.

I notice quite a few reports about dubious routing on OsmAnd's Github repo ultimately come down to OSM data being incorrect.

So, it makes a difference.
Navigation apps typically only do monthly updates of the map, so mistagging can take a while to be corrected.

“I think lit=yes can be mapped based on Mapillary if there are street lights. But maybe there could be cases where the street lights are no longer in use. I'd be happy to stay away from tagging that.”

You'd think so, but I recall the lengthy debates on StreetComplete's GH repo issues, about that assumption. It's not as straight forward as one would think.
I've encountered ways which did have lamps, but not enough to illuminate the whole way (so that their light fields overlap slightly). What's the correct value then? Unclear. Some, I've seen, have very dim lighting, or not all the lamps actually function.

If you do add lit=*, then also add something like fixme:lit=#SurveyMe and/or note:lit=* and/or source:lit=[non-survey], for those ways to be easy to find via Overpass (Turbo) to actually do a night-survey on.

Part of why I'm a stickler for quality (aside from for its own sake, principles, pride in one's work), is that I'm intending to make OSM (for Jersey) the best map (in practical terms; completeness & accuracy) as a way of drawing more users to it, and away from proprietary maps. While I know to value freedom, most others don't, and only value practical advantage. Unfortunately, almost no-one in this backward island has even heard of OSM (which sometimes leads to interesting conversations when muggles ask me what I'm doing while surveying).
To ultimately be able to say, to most any query; it's on OSM.

Even official sources for the likes of emergency=defibrillator, or amenity=charging_station are horribly inaccurate. To correct those would be a bureaucratic nightmare. But, ensuring that OSM is of high quality is quite realistic 🙂.

To then use that fact, in future, to badger / persuade officials & bureaucrats to coöperate, release data, and other goodness. I've quite a few ‘big fish’ items on my wish-list (including much better imagery, and getting the establishment to help with keeping OSM updated).

“I'm sure I missed commenting on a lot of things, but I have someone impatiently waiting for me to finish this comment so we can go.”

Understandable. Whenever you (do) have the time.

“Anyway, thanks for some really good feedback.”

It's been an interesting (& productive) conversation 🙂.

👍. Sounds like a survey is in order. Would be good to populate addr:housen* keys, too, while there.

Mentioning what was begin traced, and where, would've made it descriptive.

The likes of source & imagery_used are for indicating where data came from.

The comment is more for what, where, & why.

Much to address, here.

“The sources I've used are Mapillary, Bing and some local knowledge.”

👍

Seems that there's been some new imagery captured since I last browsed Mapillary.

“I still have some doubts about some of the tagging.”

Best to not guess. Else a tag can be misleading (compared to omitting it).

“The ones in private residential areas are the ones I am less familiar with …”

Yeah, those are tricky, and risky (see, for e.g., note/2286368 ).

“… probably also the ones that got you to raise this question, I assume.”

Not in particular, just monitoring my local area. Jersey attracts folks who just wanna pad out their How Do You Contribute? country list, or far too much osm.wiki/Tagging_for_the_renderer (or oddities like the combination of highway=pedestrian & motor_vehicle=designated on the same way) by inexperienced locals.

I'm all for quality changes.

“highway=residential”

The wiki gives a fairly clear distinction between residential versus other (service, unclassified). However, from your description, and the geometry, they do seem to qualify.

For example, a through-road which is used by non-residents, doesn't qualify. A road merely having housing doesn't make it residential (else most of town would be).

“access=private”

My reading of the wiki is that this is for when only specific individuals are permitted (e.g. an industrial area). In this case, since visitors are allowed, access=destination. Particularly since that key's value affects routing (access=private is typically treated as access=no). Housing estates marked as private aren't to be taken literally in terms of tagging; they just don't want tourists or through-traffic, which is what access=destination means.

Besides, just because signage says private, doesn't mean that this is necessarily the case, legally (think about speed limits on private land; speeding isn't then an enforceable crime, whoever put up the signage is hoping for habitual compliance). Sometimes it's used as a deterrent.

For example, if there's a public footpath at the other end, then the signage is dubious (how would a pedestrian reach the footpath?). I've seen real examples of that. Though, also of estates which very much seemed private, and had their own private footpaths connecting us surrounding roads.

However, when the only exit is also the entrance, and signage says private, then it seems more plausible.

Again; if I'm doubt, then omit (but add a fixme:highway and/or fixme:access perhaps with a value of #SurveyMe, and perhaps a note:access with a description of the ambiguity, to inform other mappers of the uncertainty; so that when someone does a survey, or happens to live there, they know that the existing values are guesswork, and to replace them).
Sounds like a survey would be a good idea, for this area, especially since ways were missing (and imagery is lacking; I often find footpaths which are missing from the map).

“ownership=private”

The entity which legally owns the land isn't determined. Thus the type of said entity isn't known. It could be a housing trust, for example, or parish holdings. Each of which warrant a different value.

In my experience, ownership is even more difficult to determine (from sources without copyright entanglements) than the operator.

“Having them tagged as service with no additional tags, doesn't seem correct to me as service roads to me means something that accesses a single house, a business, parking lots, parks, service installations (like towers).”

Usually, but it also seems to be something of a default (or used as such; perhaps incorrectly).

Even so, let's say that it was beyond question that the ways weren't service. Well, if it were also unclear what they *should* be, then guessing would be bad.
However, residential does seem apt, in this case. Simples. Though, a survey would still be good (I'll add it to my ever-growing list).

“why you think these should be service roads?”

I never said that. I think they should be tagged correctly. In this case, that seems to be highway=residential.
My caution was that I've seen non-residential (or otherwise non-qualifying) roads tagged as such.
I'm pleased to discover that you've been diligent.

Another problem in Jersey (and the other Channel Islands) is folks assuming that what applies to the UK also applies here. It often doesn't.

I do think that details (lanes, surface, smoothness, sidewalk, and so on) should be left to survey, unless unambiguously clear in imagery.
Lit almost certainly has to be by-survey.

Sometimes things like where exactly the threshold is, between two areas; where does one way become another, or where should a way be split for different tagging to apply …

Separately; yet another problem in Jersey is that we don't have enough GNSS traces to align imagery. Something else I'm working on. Thus, a lot of features are misaligned when out navigating with an app. Sometimes by enough to make it unclear which real object relates to which map element. Often, when adding a small feature, I have to choose between aligning to existing data, or to GNSS (the two are offset by at least several metres).
So, until lots more traces are available, it is wise to tag ways & building-outlines with source:geometry and (if different) source:position. This will, in future, help careful mappers realign only that which was misaligned, once enough GNSS tracks are uploaded.

Yup, it's a bit of a mess. Hence my aversion to guesswork. Often I find myself cleaning up the guesswork of others, including during precious survey time. Alas.

Since your contributions appear well-crafted; thanks for helping out.

“fictional features”

For those, use http://www.opengeofiction.net/

👍🙂

Ah, source=survey 🙂 excellent.

The technical info will be useful, too. Thanks.

“There's no trace of the hut and antenna now …”
Because they've been (re)moved?

If moved, any idea where the outer marker is, now?

Sounds like it'll make for an interesting survey trip, regardless.

Thanks. Not to worry. It was quite some time ago.

I'll report back, post-survey.

Similar to comments on changeset/116873867

In this instance a changeset comment like “Added buildings, La Rue du Craslin, Sandybrook, St. Peter, Jersey.” would've been good.

Welcome to OSM. Thanks for your contributions.

Since OSM is collaborative (and at least as much about data management as it is about producing a dataset; think about accuracy & verifiability), unambiguous changeset comments help greatly. See osm.wiki/Good_changeset_comments

Looks like you added some service roads & buildings.
In this instance, something like “Added highways & buildings near La Rue du Craslin, Sandybrook, St. Peter, Jersey.” would've been quite clear.

+1 for specifying source, and completing the tutorial.

What makes for osm.wiki/Good_changeset_comments

What makes for osm.wiki/Good_changeset_comments

Source?

👍

I happened to be passing by, recently. So, had a closer look.

The changes (from what I gather of the previous geometry) reflect reality; the minor way does join the major way, via a gate.

Ironically, however, the way (and arguably the gate, too) are very much disused. Both the way & gate are overgrown (weeds & brambles). The gate is perma-locked, rusted, disintegrating, and generally neglected.

In future, naming which road(s) or giving other location clues in your changeset comment will help local mappers.