SERVER_TYPE implies a port number (and possibly s_client options).
Previously, these were hard-coded, requiring a code change for any
new/unique services.
Now, /etc/services is used, so every assigned name is available, and
new services "just work".
The old alias names (and renames) are supported. And the old
hardcoded defaults will be used if /etc/services is not available.
SERVICES_FILE can be defined to local taste - e.g. on windows,
C:\Windows\System32\drivers\etc\services is equivalent.
FIXES#818 (I hope).
in #818, @mslavkov reported that date fails in the BG.UTF-8 locale, but
that LC_ALL=C resolved the issue.
Since we already export LANG=C, that would seem to indicate that LC_TIME
is overriding it. LC_ALL is the safer (stronger) choice.
Handle e-mail update with buggy 409 responses from registration.
Improve contact parsing by replacing call to json_get, which
doesn't seem to handle string array values well. (It's also
easier to parse the values at the same time.)
No reason to save register response JSON in TEMP_DIR,
so don't. Appears to be stale debugging code.
Exit after deactivating account.
Skip everything having to do with domains & certificates
when doing
--account-id, --new-account-key, --DEACTIVATE-account
This avoids the need to specify a domain name, creating
directories, trying to check the remote - and other unnecessary
(and sometimes harmful) work.
Most of the diffs in this commit are white space.
It's OK for the target of a CNAME not to include the source domain.
It's handy for debug and for system management.
But some people prefer a hash. We can handle that.
RFC operations for account security:
--new-account-key replaces the account key with a new one.
Can modify the type or size as well. (update .cfg first)
Does not affect certificate validity or pending operations.
--DEACTIVATE-account permanently deactivates the account on
the server. Per RFC, can not be revived. Should not revoke
existing certificates. (Server's choice.)
This is useful for debugging; it leaves any tokens in the DNS &
records its environment.
It's only meaningful for debuggers (and some problem reports
for which ask for them.
FTP_PORT not used by ftp.
No code for sftp, davfs, ftpes, or ftps.
Needs tests, but at least this won't fall thru to attempting to delete
from local file system.
Use '--value=' instead of '-v' to allow Azure CLI to process TXT record tokens that start with special characters (in particular, a dash character). Proposed fix for Issue #836.
Verifying PostgreSQL certificates requires setting the protocol type for s_client as it is not plain TLS, but behaving more like StartTLS in other protocols.