nevertheless: starting with a microcore at 3.5 mb (or less) would deserve from thin client up to other machines (i did not know that while loading tc/mc from server, you only load the tftp extension in ram, since i ever did it, but want to move to this kind of boot methode; so thanks, i have learnt smthg); curaga, you'r certainly right while saying that it is just little ram saving, but my point is rather: as we have microcore with 7.* mb, why shouldn't we make it smaller (if it is not too much workload; obviously, i don't want to complicate the existing concept)? i did not think to make for each kind of library (or group of libraries) a special extension, rather to package all libraries at the moment in microcore which we could remove into one tcz extension. if you want, it would be a sorte of what gutmensch says, i.e. radicalization of the tcz approach while at the same time a step nearer to a kind of turn key machine. as a result, you have a minimal setting with bzImage and the kernel, and everything else come as a tcz extensions. that way, you would also have an alternative to a tftp/pxe approach for those user wanting to stay with their system on a physicall support. by the way, as we are there, another suggestion (useless these are only personal thoughs in order to share ideas and stimulate the reflection on tc/mc): would it be possible for the TC-Team to provide a kind of NFS-server access in order for us to remote install tc/mc?