Terminator, but Skynet patiently explains that it is, in fact, made of infrastructure and needs working class people the way humans need gut fauna.

Perfect conditions, the wait list for a new airliner from either vendor is several years long - even with current Boeing problems. Airlines aren't canceling Boeing orders because even with money they can't replace the same order with another without increasing wait.

Now add tarriffs...

Thing is, the OSI Reference Model is not *strict ordering* and that kind of mixing was never an issue other than specific interface services (that could be provided by single protocol) logically depend on lower layers

Now, I'll agree that while most of service user operations are of the Session Layer variety, iSCSI does provide presentation layer (by virtue of simply declaring the one standard format over wire and providing encryption wrapping) and parts of application binding

... iSCSI essentially provides session layer services in supplying a transport channel for SCSI-3, including synchronization primitives - in fact, iSCSI has its own connection session as in X 215 defined as I_T nexus which spans multiple TCP connections

X.215 does concern itself with duplex, but it's not all of it. I'll agree that in a lot of cases, the L4 protocol provides big chunk of it, but that's why there's service/protocol separation in OSI. I'll agree that the login part might be wrong, but...

Because SCSI (and especially iSCSI and FC) does have a concept somewhat fitting to sessions, namely LUN logins. Once iSCSI establishes a session, you send presentation layer messages (SCSI-3) which contain application layer (specific device interfaces)

There's HyperSCSI which is SCSI transport directly on Ethernet as AFAIK mostly failed alternative to FCoE (it tried to be ATAoverEthernet but for SCSI)

Discussions of layer 5 are usually suspect, I find (after lots of relearning decades later) due to mixing up idea of "TCP session" and "session service".

iSCSI actually *does* map properly somewhat - it establishes a session service in OSI parlance, with SCSI-3 carried on top as presentation service and device-specific interfaces (like block device or tape) being application layer.

Well, scar tissue is part of "nature has healed". No scar tissue would involve (possibly divine) intervention 😉

I'd say that nature corrected it itself, given how the best practices these days seem to be CLOS networks with ethernet only as point to point medium and BGP on TOR, with ethernet L2 often being tunneled with vxlan/geneve if necessary.

Until I ended up reading somewhat more in depth documents due to #retrocomputing I never had seen for example the focus on separating Service and specific implementation aka Protocol

I'd argue that the problem with OSI is not that it is obsolete and designed for mainframes, but that somehow we started cargo culting a too-simplified diagram and ignoring everything else, then reinventing it badly.

The problem of USB-C is problem of marking and naming, and that initial physical design for the sockets was crap.

Similar how DP won me over compared to HDMI because they took better care of making sure getting right cable is easy, or how Windows-logo computers with USB4 guarantee PCIE function

WiFi brought back Ethernet's parent, ALOHA, in a way.

But expanding ethernet L2 beyond shared medium is original sin of fast ethernet and only routed ethernet (SPB/TRILL) brought good there

It stopped really being a bus with USB 2.0 hubs though.