Version 1.2.0
* libmark6sg: added automatic caching of scatter-gather block list metadata (/tmp/mark6_fuse_autocache_<hash>)
              at the first access to a scan its cache is built and it is re-used during any later accesses
* Version for DiFX-2.5, May 16, 2017

Version 1.1.9
* fuseMk6, mk6copy, m6sg_gather: use MARK6_ROOT env var if available

Version 1.1.8
* libmark6sg: simplified the prefetch method, average throughput
              now slightly better and a lot more consistent
* fuseMk6: refresh the scanlist when recordings change

Version 1.1.7
* fuseMk6 (now v1.11): fixed FUSE open() return value, error code instead of file descriptor,
           now multiple files can be opened at the same time
* fuseMk6: added initial estimate of scan size in bytes using file fragment sizes minus guessed headers
* fuseMk6: file user and group ID now match account that started fuseMk6
* libmark6sg: handle fragments where last block is only partially written out
* libmark6sg: protect file open/close internals via mutex
* libmark6sg: tidied up slightly the prefetch code

Version 1.1.6
* mark6sg/*.h: allow inclusion from C++ files

Version 1.1.5
* libmark6sg: new mark6_sg_set_rootpattern() API function
* fuseMk6 (v1.10): new command line option '-r <rootpattern>',
*          for example -r /mnt/disks/[1-4]/[0-7]/data/
* support out-of-tree builds (TODO: use static linking to libjsmn)

Version 1.1.0
* libmark6sg: new mark6_sg_stripesize() API function
* libmark6sg: blocklist extraction with one thread per file for major speedup
* libmark6sg: read functionality reimplemented via mmap()
* libmark6sg: added background threads that touch upcoming mmap()'ed pages,
              causing kernel to prefetch the pages in parallel from all files
              and giving factor ~10 higher throughput (now ~1 GB/s)
* fuseMk6: new command line option -v
* mk6copy: simple copying utility like 'm6sg_gather' but with timing reports

Version 1.0.0
* Support the Mark6 Scatter-Gather formats v1 and v2