Reply 20 of 21, by wierd_w
The latency issue crops up in combination with chattyness and multiple devices in a collision domain.
It just gets greatly magnified over WAN links.
The WAN part of the quote is not the impirtant part anyway. The 64kb max blocksize is.
For most small files of the dos era, this means a single message can contain the entire file, and local caching works just peachy.
The 'do not do this ever!' Admonition, is
'Do not try to mount an .iso or other large disk image directly from an smb1.0 share; the protocol is designed for small files, has problems with cache concurrency without oplocks, saturates the link with protocol control messages and small nibbled datagrams on reads or writes to large files, and hangs up if a transaction 'takes too long', all of which contributes to timeouts and orphaning of the file handle.
DOS block device drivers are not equipped to recover from orphaned/stale handles, and will just give read errors once that happens until you unmount, then remount the image. from their perspective, you drive is having a general read error/damaged.
If you want to work with image files over the lan, you should use a protocol that is better suited.'