what is strange is that using the raster way i can see all the item available in the server but can't have access to the drawing layer on the server !!!
inspire.cadastre.gouv.fr has many server configuration errors. It is one of the test cases Manifold engineering has used to develop technical means in Future to get around server errors.
I don't remember if it was cadastre specifically or some other server (US government servers are often truly awful as well... NOAA comes to mind...) but a typical server error is when the server declares image tiles that it serves as being text and not images. If you try to connect to that server using a WMS client which honestly applies WMS standards, OGC says that is not supposed to work and with Release 8 it will not work. It doesn't matter that the layer is listed as it will not appear due to the server misconfiguration.
Dealing with such server errors is not so simple, because however you do it you will anger some people. It is not so simple as just applying standards accurately and honestly. There is also a legitimate desire to extract all information possible even from broken data sources.
Release 8, for example, comes from a culture in which there was a greater focus on the value of standards. It was understood by all that ignoring the details of standards debased those standards, and that debasing standards in turn would eventually deny everyone the benefits those standards provided.
In that world, applying WMS standards correctly was praiseworthy and, instead of demanding that Release 8 be debased to lower IQ, folks would have focused their efforts on correcting the errors in the cadastre.gouv.fr server.
That approach doesn't work in the broader world of today, where the broadening of access to GIS, a good thing, has had the unfortunate side effect of not being accompanied by broadening of insight and education. So now, if people instantly don't get what they want they blame whatever is closest at hand, even if that is the thing which is functioning correctly.
Radian therefore automatically switches into slacker mode as necessary when connecting to the alphabet soup of web server "standards". It is a bit like how over the years Release 8 collected broken shapefiles, evolving the ability to decode a very wide "zoo" of files that claimed to be shapefiles but which violated the shapefile spec in all sorts of bizarre ways.
Radian is steadily building a zoo of configuration errors in web servers and is gaining greater and greater ability to guess when a server is making one of those errors and thus that error should be ignored.
There is a lot of risk in that approach, of course, since applying guesswork is necessarily imperfect. But that's life. You get no gold medals for honoring standards, but you get plenty of praise if you just connect to zillions of servers no matter how misconfigured those servers may be.
To get around possible errors in cadastre.gouv.fr, try the latest edition of Future and see if it connects. Current Future builds already include many workarounds to server errors. The latest build (probably coming out today or tomorrow) might include even more.
If you ever find a "WMS" site (or one of the other alphabet soup web server standards) that the latest version of Future cannot use, please send in a bug report to tech support as discussed in the Bug Reports page. Engineering can then investigate and possibly expand the zoo of configuration errors for which there are workarounds.