[Geoserver-devel] Are there png8 wms tile problems with IE?

I am cross posting this on both geoserver & openlayers lists..

Recently I was enquiring if any of you have seen the exception issues with geoserver 1.5.3. I have further problems with 1.5.3, this time with PNG8.

Is the png8 format rendered by geoserver-1.5.3 fully compliant with the PNG specs? Please ignore my lack of knowledge on PNG. Are there any limitations or certain features not being implemented for this png8 format? I am using Openalyers 2.4 on the client side. Everything is fine on firefox. But both IE6 & IE7 have problems with the png8 produced by geoserver. This manifests in couple of ways:

1) Most of the times, the IE gets redirected to some other page when I use png8 format for openlayers wms layer. (If I use png instead of png8, everything is fine). I am hypothesizing that IE is not able to process the geoserver png8 and gets confused.

2) Lot of times with png8, the openlayers lands in OpenLayers.Util.onImageLoadError() handler. Basically IE thinks that there is some problem with image URL; however the IE ends up drawing the WMS content of the png8 tile fine (but the tile soon gets replaced by a pink tile). Looks like IE thinks that there is something wrong with this tile even though it was able to render it fine. The problem does not consistently happen with one particular tile; the same tile on refresh comes out fine.

The support for png8 is very very important for IE6 since IE6 can't handle PNG24 transparency.

Any body has any ideas? Thanks in advance..
Louvy Joseph

Hi Louvy,

I’ve noticed that the png8’s had a problem in several pieces of software… not just IE. I suspect (and I do stress suspect) that this may be caused not by a defective image or the inability of IE/Others to handle 8 bit png… but I think it has something to do with the HTTP response type of the image from geoserver. Correct me if I am wrong… but “image/png8” is not an official mime-type, its likely IE is getting confused by this. PNG was designed as a wrapper to several compression formats… hence the name Portable Network Graphics.

In IE, I noticed that I was getting black instead of transparency in the image inside IE. Perhaps this is the same for you?

In the end, rather than use png (24) or png8 I switched to gif.

–AH

On 9/21/07, louvy.joseph@anonymised.com <louvy.joseph@anonymised.com> wrote:

I am cross posting this on both geoserver & openlayers lists…

Recently I was enquiring if any of you have seen the exception issues
with geoserver 1.5.3. I have further problems with 1.5.3, this time with
PNG8.

Is the png8 format rendered by geoserver-1.5.3 fully compliant with the
PNG specs? Please ignore my lack of knowledge on PNG. Are there any
limitations or certain features not being implemented for this png8
format? I am using Openalyers 2.4 on the client side. Everything is fine
on firefox. But both IE6 & IE7 have problems with the png8 produced by
geoserver. This manifests in couple of ways:

  1. Most of the times, the IE gets redirected to some other page when I
    use png8 format for openlayers wms layer. (If I use png instead of png8,
    everything is fine). I am hypothesizing that IE is not able to process
    the geoserver png8 and gets confused.

  2. Lot of times with png8, the openlayers lands in
    OpenLayers.Util.onImageLoadError() handler. Basically IE thinks that
    there is some problem with image URL; however the IE ends up drawing the
    WMS content of the png8 tile fine (but the tile soon gets replaced by a
    pink tile). Looks like IE thinks that there is something wrong with this
    tile even though it was able to render it fine. The problem does not
    consistently happen with one particular tile; the same tile on refresh
    comes out fine.

The support for png8 is very very important for IE6 since IE6 can’t
handle PNG24 transparency.

Any body has any ideas? Thanks in advance…
Louvy Joseph


Dev mailing list
Dev@anonymised.com
http://openlayers.org/mailman/listinfo/dev

Andrew Hughes ha scritto:

Hi Louvy,

I've noticed that the png8's had a problem in several pieces of software... not just IE. I suspect (and I do stress suspect) that this may be caused not by a defective image or the inability of IE/Others to handle 8 bit png.. but I think it has something to do with the HTTP response type of the image from geoserver. Correct me if I am wrong... but "image/png8" is not an official mime-type, its likely IE is getting confused by this. PNG was designed as a wrapper to several compression formats... hence the name Portable Network Graphics.

The mime type returned by the png8 map producer is still "image/png".

In IE, I noticed that I was getting black instead of transparency in the image inside IE. Perhaps this is the same for you?

In the end, rather than use png (24) or png8 I switched to gif.

Just tried the following request:
http://localhost:8080/geoserver/wms?HEIGHT=550&WIDTH=429&LAYERS=tiger%3Apoly_landmarks&STYLES=&SRS=EPSG%3A4326&FORMAT=image%2Fpng8&SERVICE=WMS&VERSION=1.1.1&REQUEST=GetMap&EXCEPTIONS=application%2Fvnd.ogc.se_inimage&BBOX=-74.07079%2C40.661263%2C-73.884215%2C40.900463&transparent=true
in both firefox and ie7/win xp and ie7 Vista, works fine
in both, no black areas here...
I don't have access to and older IE unfortunately...

Cheers
Andrea

louvy.joseph@anonymised.com ha scritto:

I am cross posting this on both geoserver & openlayers lists..

Recently I was enquiring if any of you have seen the exception issues with geoserver 1.5.3. I have further problems with 1.5.3, this time with PNG8.

Is the png8 format rendered by geoserver-1.5.3 fully compliant with the PNG specs? Please ignore my lack of knowledge on PNG. Are there any limitations or certain features not being implemented for this png8 format? I am using Openalyers 2.4 on the client side. Everything is fine on firefox. But both IE6 & IE7 have problems with the png8 produced by geoserver. This manifests in couple of ways:

1) Most of the times, the IE gets redirected to some other page when I use png8 format for openlayers wms layer. (If I use png instead of png8, everything is fine). I am hypothesizing that IE is not able to process the geoserver png8 and gets confused.

2) Lot of times with png8, the openlayers lands in OpenLayers.Util.onImageLoadError() handler. Basically IE thinks that there is some problem with image URL; however the IE ends up drawing the WMS content of the png8 tile fine (but the tile soon gets replaced by a pink tile). Looks like IE thinks that there is something wrong with this tile even though it was able to render it fine. The problem does not consistently happen with one particular tile; the same tile on refresh comes out fine.

I would be happy to investigate this, but I can't reproduce any of these
problems with the geoserver standard data set. If you can give me
a full set of data, requests and whatnot needed to reproduce these
errors I'll be happy to have a look at it.

Cheers
Andrea

Thanks Andrea & Adrew for tips.

I think, I may have fixed the browser redirect issue. Sad to say that the quality of geoserver-1.5.3 is not good. We have another war in the same tomcat container. Looks like geoserver is getting into some kind of trouble and then pushes our main app in the same container over the edge. And that was resulting in browser redirects. When we separate the geoserver-1.5.3 & our app into separate tomcat containers, everything was fine. Similar issue was not found with geoserver1.5.0 and neither with geoserver1.6-beta3. I do see tonnes of exceptions in the tomcat logs from geoserver-1.5.3 but it seems to recover by itself when we isolate it into its own tomcat container.

The occassional png8 with IE thinking it as bad format and resulting in pink tiles in openlayers is not solved yet. I don't have shareable dataset. It is a huge database. Let me investigate further and I will feed questions/info to the mailing list.

BTW, the 1.6beta3 is very very fast. That is a good news. (But we have some other issues with wms labels not appearing and some wfs commits failing. I need to find out more but I would like to move 1.6 soon). What are the 1.6 time lines?

Thanks again for wonderful support
Louvy

Andrea Aime wrote:

louvy.joseph@anonymised.com ha scritto:

I am cross posting this on both geoserver & openlayers lists..

Recently I was enquiring if any of you have seen the exception issues with geoserver 1.5.3. I have further problems with 1.5.3, this time with PNG8.

Is the png8 format rendered by geoserver-1.5.3 fully compliant with the PNG specs? Please ignore my lack of knowledge on PNG. Are there any limitations or certain features not being implemented for this png8 format? I am using Openalyers 2.4 on the client side. Everything is fine on firefox. But both IE6 & IE7 have problems with the png8 produced by geoserver. This manifests in couple of ways:

1) Most of the times, the IE gets redirected to some other page when I use png8 format for openlayers wms layer. (If I use png instead of png8, everything is fine). I am hypothesizing that IE is not able to process the geoserver png8 and gets confused.

2) Lot of times with png8, the openlayers lands in OpenLayers.Util.onImageLoadError() handler. Basically IE thinks that there is some problem with image URL; however the IE ends up drawing the WMS content of the png8 tile fine (but the tile soon gets replaced by a pink tile). Looks like IE thinks that there is something wrong with this tile even though it was able to render it fine. The problem does not consistently happen with one particular tile; the same tile on refresh comes out fine.

I would be happy to investigate this, but I can't reproduce any of these
problems with the geoserver standard data set. If you can give me
a full set of data, requests and whatnot needed to reproduce these
errors I'll be happy to have a look at it.

Cheers
Andrea

louvy.joseph@anonymised.com wrote:

Thanks Andrea & Adrew for tips.

BTW, the 1.6beta3 is very very fast. That is a good news. (But we have some other issues with wms labels not appearing and some wfs commits failing. I need to find out more but I would like to move 1.6 soon). What are the 1.6 time lines?
  

Thanks for the feedback, with stress tests like you are providing GeoServer 1.6 is easier to get out the door. Everyone is going to be away next week at FOSS4G (just in case you are wondering why email and IRC are so quiet).

Thanks again for wonderful support
  

I agreed - the TOPP team has done really great work for 1.6. Both in terms of WFS1.1 and flat out speed.

I should remind everyone that The Open Planning Project team are available for hire (so if those "other" issues are holding you up why not talk to them?).

Cheers,
Jody

On Sat, September 22, 2007 7:21 am, louvy.joseph@anonymised.com wrote:

Thanks Andrea & Adrew for tips.

I think, I may have fixed the browser redirect issue. Sad to say that
the quality of geoserver-1.5.3 is not good. We have another war in the same
tomcat container. Looks like geoserver is getting into some kind of
trouble and then pushes our main app in the same container over the edge.
And that was resulting in browser redirects. When we separate the
geoserver-1.5.3 & our app into separate tomcat containers, everything was
fine. Similar issue was not found with geoserver1.5.0 and neither with
geoserver1.6-beta3. I do see tonnes of exceptions in the tomcat logs from
geoserver-1.5.3 but it seems to recover by itself when we isolate it into
its own tomcat container.

I'm surprised you don't see the same problem with GeoServefr 1.6 since
it depends on even more libraries, so the permanent generation problem
should be even worse. But this problem shows up only after some time,
as the VM loads more and more classes as it needs them.
Anyways, if you add -XX:MaxPermSize=128m to your VM configuration
you should be ok, see also
http://www.brokenbuild.com/blog/2006/08/04/java-jvm-gc-permgen-and-memory-options/

The occassional png8 with IE thinking it as bad format and resulting in
pink tiles in openlayers is not solved yet. I don't have shareable dataset.
It is a huge database. Let me investigate further and I will
feed questions/info to the mailing list.

BTW, the 1.6beta3 is very very fast. That is a good news. (But we have
some other issues with wms labels not appearing and some wfs commits
failing. I need to find out more but I would like to move 1.6 soon). What
are the 1.6 time lines?

I tried to reproduce your issues on a Vista box but failed... I really
need some sample data and SLD to figure out what's going on...

Cheers
Andrea

aaime wrote:

Just tried the following request:
http://localhost:8080/geoserver/wms?HEIGHT=550&WIDTH=429&LAYERS=tiger%3Apoly_landmarks&STYLES=&SRS=EPSG%3A4326&FORMAT=image%2Fpng8&SERVICE=WMS&VERSION=1.1.1&REQUEST=GetMap&EXCEPTIONS=application%2Fvnd.ogc.se_inimage&BBOX=-74.07079%2C40.661263%2C-73.884215%2C40.900463&transparent=true
in both firefox and ie7/win xp and ie7 Vista, works fine
in both, no black areas here...
I don't have access to and older IE unfortunately...

Cheers
Andrea

I've just moved to another machine which uses IE6 and I'm having similar
problems, but it works fine in Firefox and IE7. My tiles usually end up
having a grey background instead of transparency, but sometimes it all goes
pink. I'm using gifs now but this seems slower.
Jon

--
View this message in context: http://www.nabble.com/Are-there-png8-wms-tile-problems-with-IE--tf4496219.html#a13135403
Sent from the GeoServer - Dev mailing list archive at Nabble.com.