This might be too far off the original idea... But: It might be
interesting to look into a more disconnected way of "managing"
geospatial data with the Web. As in having "servers" advertize latest
changes in geometry and attributes using GeoRSS and "clients" picking
these up and integrating them (potentially including processing,
changes) into their "cache" or "copy" of the data. Which might
automagically turn into a "server" itself advertizing changes following
the same pattern. This idea grew on me over the past years after seeing
Ward Cunningham give his presentation at Wikimania 2005 in Fankfurt on
bits of information floating around the Web [1]. Maybe now the time is
ready to implement some of these thoughts in our domain.
I am currently seeding these ideas to the European National Mapping
Agencies through the ESDIN project. Yes I know, it sounds a bit remote
but hey, it is a nice vision... :-)
Have fun,
Arnulf.
[1] Slides: http://c2.com/doc/wikimania/
Screener: http://www.archive.org/details/WIKIMANIA_TAPE_019_SCREENER.mov
Low res video runs 30 minutes
On 09/24/2010 01:40 AM, Bruce Bannerman wrote:
Ragi,
I agree. I think that we have a way to go yet to have something
comparable to the ArcSDE / ArcGIS Multi-versioning and version conflict
detection functionality.
The advantage that the ArcSDE solution has is that edits are made
directly within the database. This works well within an Enterprise
environment as described by Fabio earlier in this thread.
I may be wrong, but I think that git works on files, but I haven’t used
it myself. Can git detect changes to the spatial representation of a
feature within a binary file?
Also, speaking as someone who implemented an ArcSDE/ArcGIS
Multi-versioned edit scenario several years ago, the ESRI solution is
far from perfect. It imposes very strict environment management on the
system managers, e.g.:
* All versions of the software used (client and server) must be at
precisely the same version, service pack and patch;
* The environment can only use software that implements the
ArcObjects environment (from experience, this rules out the use of
the ArcSDE Java and C API’s);
* Editors must be well trained and knowledgeable in using both
ArcGIS and Multi-versioned processes;
* The Organisation needs to think through their maintenance
processes to get best advantage of the functionality; and
* It doesn’t remove the need for data maintenance people to talk to
each other about work that is going on, as the software cannot
resolve all conflicts. For example, if two editors make changes to
the spatial representation of a feature, which one is correct? The
software will detect the conflict, but the editors (or their
managers) will need to resolve the issue of which version of the
feature’s spatial representation is correct.
Bruce
On 24/09/10 4:05 AM, "Ragi Burhum" <r...@burhum.com> wrote:
Hi Noli,
thanks for the link. That is definitely a step in the right
direction, but it is hardly comparable to git ArcSDE versioning at that.
The article and sample code you describe above generates hashes for
all rows and tables in the db and compares them to the target db. So
1 million rows in a db, regardless if the two dbs are identical,
would cause 1 million hashes to go over the wire. Every single time
you ask to sync you pay the price.
Git and ArcSDE keep track of changesets, and when it is time to
synchronize, they exchange that changeset and apply it. One insert?
That is all that needs to be sent.
Another issue is that there is nothing about conflict resolution
there (what happens when you delete one row in one db and modify it
in another one?). There is also the problem of allowing multiple
versions of the data in the same db (Like having multiple heads).
Regardless, thank you for the link,
- Ragi
Date: Thu, 23 Sep 2010 13:22:17 +1000
From: Noli Sicad <nsi...@gmail.com>
Subject: Re: [OSGeo-Discuss] Re: "git" like for geodata management
To: OSGeo Discussions <discuss@lists.osgeo.org>
Message-ID:
<aanlkti=3anc4baand4hk9uuzfsasxn-8ybpnkyong...@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
PostgreSQL Synchronization Tool --- psync [1]
" The article introduces a method of synchronizing two PostgreSQL
databases. Although, this seems to be an easy task, no product
(slony,
londiste, ...) really satisfied the needs within the
maps.bremen.de <http://maps.bremen.de> <http://maps.bremen.de>
project. Either they have special prerequsits that didn't apply for
our problem or they didn't support synchronizing of large objects.
Large objects are used to store tiles of a street/aerial map within
PostgreSQL. My GIS-server queries the database and gets the
tiles out.
By using this construction we are getting a flexible infrastructure
for updating and maintaining different versions of the maps.
Everything was working fine until the service needs to be spread
over
three servers. How can we easily synchronize the databases? I really
found no really working solution that is clean and easy to use. "
[1]http://www.codeproject.com/KB/database/psync.aspx
<http://www.codeproject.com/KB/database/psync.aspx>
Noli
On 9/23/10, Ragi Burhum <r...@burhum.com> wrote:
Are you looking for an alternative to (1)ESRI's versioning,
(2)ESRI's
disconnected editing, or a mix of both (3)git like? the
scenario that you
described first was more like (2), but this one fits (1).
I would love to see something like (3), but truth of the
matter, AFAIK,
there is nothing like that implemented for geo (yet).
On Sep 22, 2010, at 9:00 AM, discuss-requ...@lists.osgeo.org
wrote:
On Wed, 2010-09-22 at 12:10 +0800, maning sambale wrote:
Any real world cases for this?
Imagine the following scenario:
* 50 ~ 70 digitizers
* 5 QA
* 1 Manager
Each QA has 10 digitizers assigned. After all the data
is validated, the
manager merges it and generates the geodb.
All users work against the same DB, most of them linked.
This causes
disconnections, duplicated data, and lots of random errors.
Also, they can't be forced to work on different DB's
because they are
all working on the same project, at the same time.
This is the real scenario of GISWorking
(http://www.gisworking.com/), a
company we are working with.
It would be perfect to have smaller groups (ideally 1
person), working
against separated databases, but that can be
synchronized with the rest
of the data when needed.
Then each QA merges data from the people he supervises.
After it's
validated the manager merges the complete dataset, and
generates the
final "product".
I don't know if this it's the exact same case, but we
are working on it
with a similar approach.
_______________________________________________
_______________________________________________
Discuss mailing list
Discuss@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/discuss
--
Exploring Space, Time and Mind
http://arnulf.us
_______________________________________________
Discuss mailing list
Discuss@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/discuss