Jean-Marie, A minor point is that you don't need to specify any url attribute in a dataset tag if you are going to specify them in the variable attributes. So your dataset tag could look like this: <EC1_Historic_time_nc name="Simulation run 1 with green-house gas">Of course we always encourage people to have documentation links but that's not required either. -- Jon J-M Epitalon wrote: Adam, there is a third solution: LAS provides also for another way of creating virtual datasets (beside categories) In an XML datset description file, you can redefine for each variable the actual filename that contains it; See the following example: <datasets> <EC1_Historic_time_nc name="Simulation run 1 with green-house gas" url="" class="moz-txt-link-rfc2396E" href="">"file:dummy.nc" doc=""> <variables> <!-- Below, the attribute "url" allows for defining the actual file path and name --> <CORESERV name="RESERVOIR " url="" class="moz-txt-link-rfc2396E" href="">"file:/home/glo_data/Discendo/Europe/1950-1999/EC1/EC1_1m_195001_199912_CORESERV.nc"> <link match="/lasdata/grids/Medias_flat_monthly_1950_1999_grid"/> </CORESERV> <PLDETESU name="TEMPERATURE " url="" class="moz-txt-link-rfc2396E" href="">"file:/home/glo_data/Discendo/Europe/1950-1999/EC1/EC1_1m_195001_199912_PLDETESU.nc"> <link match="/lasdata/grids/Medias_flat_monthly_1950_1999_grid"/> </PLDETESU> Then, variables TEMPERATURE and RESERVOIR seem to be in the same dataset (file): EC1_Historic_time_nc that appears under the name="Simulation run 1 with green-house gas" Hope this helps Jean-Marie CERFACS Toulouse, France On Tue, 2005-03-22 at 17:10, Roland Schweitzer wrote:Adam, Adam Baxter wrote:Thanks Everyone! It works great! I really appreciate all the help I've been given. One final question and I should be set. We're separating the control files by variable for the moment, mostly due to system constraints . From what I've seen with the xml, each dataset usually only has one control file to point to. Is this correct? The documentation has some information on creating categories - would this be of any use to coalesce the variables under one category, even if they're under different datasets due to having only one url per dataset? AdamAt the risk of confusing the situation, let me offer a few comments. The concept of a "dataset" is nebulous at best. (Personally, I don't even like the term since it's not even an English word, but that's another prejudice all together). It seems to me in the world of OPeNDAP a single URL is called a dataset regardless of what it contains (one variable or 20 variables). And if I understand your situation correctly, each of your OPeNDAP URL's points to an aggregation (accomplished via a GrADS control file). It's likely therefore that the process by which you created your LAS XML resulted in a pretty flat presentation of the underlying data. You likely have LAS XML that looks like: <datasets> <ds1 name="physical parameter1" url="" class="moz-txt-link-rfc2396E" href="">"http://gds.gov/urlpp1"> <variables> <pp1 name="physical parameter 1"> <link match="/lasdata/grids/grid1"> </pp1> </variables> </ds1> </datasets> <datasets> <ds2 name="physical parameter2" url="" class="moz-txt-link-rfc2396E" href="">"http://gds.gov/urlpp2"> <variables> <pp2 name="physical parameter 2"> <link match="/lasdata/grids/grid1"> </pp2> </variables> </ds2> </datasets> In LAS, the term dataset is used to describe a container that has "variables" as children. A "category" on the other hand is a container that has categories or datasets as children. What this means in practical terms, is that when you click on a dataset you get a list of variables each having a "check box" when allows you to select one or more to use. Datasets and categories appears as "links" (not check box items) and when you click on one you get more links (in the case of categories) or a list of variables with check-boxes (in the case of a dataset). Without using <las_categories> your presentation will always be a two layer hierarchy. You'll see a page of links and when you click on the link and you see a list of variables, but in your case it's likely that you list of variables each only contains 1 item (if your XML looks like the above). If I were faced with this situation, the first thing I would do is reorganize my XML to look like this: <datasets> <ds1 name="some interesting data" > <variables> <pp1 name="physical parameter 1" url="" class="moz-txt-link-rfc2396E" href="">"http://gds.gov/urlpp1"> <link match="/lasdata/grids/grid1"> </pp1> <pp2 name="physical parameter 2" url="" class="moz-txt-link-rfc2396E" href="">"http://gds.gov/urlpp2"> <link match="/lasdata/grids/grid1"> </pp2> </variables> </ds1> </datasets> Where related variables are grouped together (each with their own data url) instead of each being in a separate dataset with the url on the dataset element. Since you have a large collection you might want to invest the time to write a script that can do this manipulation for you. Then if you're satisfied with the two level presentation of the data that results, you're done. If you want to further group data into categories to separate say the "surface" data form the "upper atmosphere" data then you can start introducing categories. I hope this helps. Roland |