[OSM-talk] Proposal for import guidelines

[Paul Norman]

My main machine is down at the moment so this isn't as detailed as I'd like, but I have a few thoughts.

On Sep 25, 2012, at 10:11 AM, Richard Fairhurst <richard at systemeD.net> wrote:

A propos of the recent contretemps about Cadastre imports and separate 
accounts (excessive use of French in this sentence is unintentional), 
I'd like to propose the following modification to the import/bulk edit 
guidelines:

==

An 'automated edit' is one where the editing is not carried out by 
manual drawing actions. This includes (but is not limited to):

- imports of external data
- search-and-replace tag changes
- automated geometry fixup
- reverting edits

and applies equally to scripted edits and to those carried out within an 
editor program.

All changesets including automated edits MUST have the following 
additional tags:

bot=yes
bot_url=<link to a page describing the automated edit>

Users are also encouraged to add these tags:

bot_type=<machine-readable description of the edit type>
bot_source_licence=<machine-readable licence name>

For example, bot_type=import, bot_source_licence=public_domain; or 
bot_type=revert.

The tags should be added to the changeset, not the individual objects. 
Authors of software facilitating such edits (e.g. editor plugins) should 
provide relevant tags as a default.
 
I'd like to see a standardized tag to indicate reverted changesets. The redaction bot used redacted_changesets=*, perhaps reverted_changesets=cs1;cs2;cs3 (etc) could be used as a standard way to indicate changesets

In addition, all automated edits of a high-volume, sustained or 
continuous nature MUST also be carried out from a separate OSM account. 
This includes (but is not limited to):

- large-scale imports (for example, 20,000 nodes or greater)
- continuously running scripts
- edits affecting more than one country
 
Although I agree that these should or do require a separate account, I wouldn't classify large-scale imports as automated edits, I'd classify them both as types of bulk edits.

I see bulk edits as falling into two groups

- Mechanical edits, some of which would be automated edits

- Imports

In theory you can have edits which blur the boundaries (e.g. recent Czech edits) but in practice these are infrequent. Most bulk edits clearly fall into one group or another.
 

Like all other mappers, authors of automated edits must monitor the OSM 
inbox for any accounts they use, and be prepared to respond to messages 
and queries about their edits.
 
Speaking as someone who both maintains multiple (four) accounts and frequently has to contact separate accounts, I prefer a link to the person's main account, either to /user/name or /message/user/new. In theory all messages sent to any of my accounts result in an email to my main account but in practice I've found these sometimes get routed to spam. I regularly check my main account but I normally only check the others on demand. Even if I was checking these daily my main account has more detailed contact information and information on how to get in touch with me quickly.

We recognise that complying with this rule may seem onerous, but we 
would remind authors of automated edits that "with great power comes 
great responsibility". OpenStreetMap's value, and differentiation from 
other data providers, comes from the local knowledge, skill and 
enthusiasm of its community, rather than from simply agglomerating data 
available elsewhere. These guidelines are designed to retain visibility 
of automated edits and thereby safeguard our most precious resource.

==

(end of proposed text)

I hope you can see the intentions behind this proposal, but in essence:

- requiring particular tags makes visibility easier, so that DWG et al 
have a better view of automated edits;
- it also helps to spread awareness of automated edits through the 
community, since these edits can be easily visualised by client software 
- thereby bringing "many eyeballs" to the edits;
- encouraging a machine-readable licence tag helps to avoid the issues 
identifying changesets that were encountered in the redaction.

A brief clarification on this message: This is a personal posting. I 
have already proposed to the OSMF board that the three similar sets of 
guidelines on the wiki (imports, automated edits, mechanical edits) be 
combined into one, and that the result is endorsed as an OSMF policy. If 
this suggestion is received reasonably positively, then I'll bring it 
forward for incorporation into such a policy.

I would welcome your comments. :)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20120925/70f95bed/attachment.html>