Hi, Wolfgang Solfrank: > I'm a bit scarce on round tuits right now,
The global tuits shortage crisis is of course a big obstacle. Especially since ISO 9660 is essential when booting installation CDs or DVDs. I plan to repack NetBSD-6.1.3-i386.iso with my development kernel and to perform an installation. (It is not nice of makefs to hide the El Torito boot image in blind space. No i have to guess its size.) Currently i am looking for critical comments and questions about the overall implementation concept. A decision would be nice, whether API compatibility for cd9660_node.h is worth an extra pointer in each iso_node (= vnode.v_data). As stated, fstat(1) would be affected by an API break. On the other hand, half of my CVS /usr/src seems to suffer from API problems with the NetBSD-6.1.3 system installation. Heavy testing makes sense only after this decision. > RRIP doesn't have the concept of multiple versions of a file. Yes. It looks funny if there is more than one ISO file version which all have the same RRIP name. I tested this. My proposal showed the best results among the tested systems. > What 6.1.3 do you refer to here? NetBSD-6.1.3 i386. The host operating system is Debian 6 amd64. My development subject from CVS refers to itself as NetBSD-6.99.40 (#190 meanwhile). > Regarding associated files, I saw those used by Apple for their > resource fork. Albeit I haven't seen any ISOs with them recently. Probably i will have to fake some by hand. This implies the risk, though, that i got the specs wrong in both cases. A real test case would be much more significant. In general i would say that large files are much more common than Associated Files, multiple Versions, or ISO 9660 Extended Attributes. So even if there is a risk to break those, it still seems worth. Have a nice day :) Thomas
