Cloud storage

NIST SP 800-145 defines cloud computing as on-demand network access to a shared pool of resources. Storage is one of those resources. The phrase does not name a protocol. The transfer into it is still HTTPS, SFTP, or whatever the provider accepts.

Two shapes get the same label. Object storage, the S3 pattern, stores opaque objects by key. A file service, the drive-mount pattern, stores a hierarchy you can lock and edit in place. Sync folders are a third shape: a client keeps a local copy and reconciles it with the service. A 10 GB upload into a bucket, a mapped drive save, and a background sync of the same 10 GB are three different failure modes wearing one product name.

A team "puts the project in the cloud" by dropping a folder into a sync client. The client uploads changed chunks overnight. A colleague on a plane edits the same deck. Both copies exist until the client asks a human to pick one. That is sync, not a completed file transfer. The same team later sends a client a bucket link to a finished export. That export is an object, versioned or not, with no merge. Calling both cloud storage is fine. Treating them as the same handoff is how a merged-wrong deck ships.

Cloud storage is not a residency guarantee and not an encryption guarantee. Region is a bucket setting. Keys are a separate setting. The provider can usually read server-side-encrypted objects. Egress fees are the bill for downloads, which makes a CDN or a same-region compute client a cost decision rather than a protocol decision. "Unlimited cloud storage" in a consumer plan is a fair-use policy. It is not a transfer mechanism.