jmscott / blobio Goto Github PK

A typo in the BLOBIO_SERVICE yields a "blob not found " exit status, which
is confusing. Instead, the existence of the data directory ought to
tested for a more clear error message. An earlier version of fs did
just this test.

for example, the sha_fs using only the final 9 ascii characters. should, instead, the final file name be the full digest?

so the path sha_fs/0/0d/8364/84e19434/70efa4498a72c5d7/e1c9f4c73 becomes
sha_fs/0/0d/8364/84e19434/70efa4498a72c5d7/e1c9f4c73/00d836484e1943470efa4498a72c5d7e1c9f4c73

when exit_status is not ok then the value becomes null in when qualifications, which makes fault handling awkward.

what's the difference between qualifying on null and the the rule simply never firing. tricky.

in the blobio c client the rofs service hard links to the blobio file, eliminating the need for a copy. unfortunately, this implies the caller can't remove the link unless the caller owns the parent directory.

trusting would only apply to get and put verbs.

should trusting be also triggered as a syntax extension to the service?

systemd reports unusually higher memory usage for group of biod processes. real memory is low, according to top and atop, so the existence of the possible leak is not clear. a possible source of the leak may be the wrap command for in brr logger.

When you will sell blob.io domain to a more alive project?

for example, the following code breaks in wierd ways:

    sql query abc row {
        result row is (is_xml bool)
        ...
    command is_xml { ...

An unqualified "when" clause should always evaluate to true, implying the rule always fires. Unfortunately, flowd will periodically not fire the unqualified rule. Clearly this is a bug. Fortunately, the bug is reproducible under both Mac OSX and Linux. I've only seen the bug for unqualified rules, which are always root nodes of the directed process graph.

the syntax would model postgres. BLOBIO_SERVICE=/usr/local/blobio/data.

An option for tail to create a non-existent file, as specified in "path" variable, would be helpful.
For example, inbound spool/prefixio.brr files may not exist until first brr is sent, which cause tail to panic.

The current fix is touch spool/prefixio.brr on startup when etc/profile is source by boot-flowd.

disabled commands would always evaluate to null.

KISS.

currently 'blobio eat --algorithm --input-path" only writes the digest, which becomes a pain in shell scripts.

tail ought to log the percentage of the current log file that has been scanned.