Identifiable

Generates GUIDs based on current time (UUIDv1), or based on a unique name using a cryptographic hash (UUIDv3 using MD5 or UUIDv5 using SHA1). Includes non-standard variant layouts of time-based GUIDs for near-sequential sorting in SQL Server and PostgreSQL.

TimeGuid.Create

Generates time-based GUIDs for use in database clustered indexes. The following algorithms are implemented:

• Standard generates a time-based GUID that conforms to RFC 4122 without exposing a MAC address.
• SqlServer generates a similar identifier that is optimally arranged for use with SQL Server clustered indexes and can be transposed to/from a valid Version 1 identifier. In this arrangement, the variant will be zero to avoid mis-identification as an invalid GUID version.
• PostgreSql generates a similar identifier that is optimally arranged for use with PostgreSQL clustered indexes and can be transposed to/from a valid Version 1 identifier. In this arrangement, the variant will be zero to avoid mis-identification as an invalid GUID version. Why not use a random GUID? Random GUIDs have well-known performance issues when used in clustered indexes. These issues have been mitigated by the availability of faster hardware, but the performance of random GUIDs is still significantly worse than time-based GUIDs. The inability to sort by time is also a significant drawback.

Why not use NEWSEQUENTIALID or UuidCreateSequential?

• NEWSEQUENTIALID is only usable as a default constraint and the values it generates are not a valid GUID version (bytes 6/7 are reversed, causing the version to be in the wrong position).
• UuidCreateSequential requires P/Invoke and is only available on Windows.
• UuidCreateSequential and NEWSEQUENTIALID are only sequential when generated from the same system (the time component is in the least-significant bytes for SQL Server sorting).
• UuidCreateSequential and NEWSEQUENTIALID both include the MAC address which may be undesirable or non-unique (a pervasive problem with virtual machines).

Why not use a COMB? The values generated by this method are, in fact, COMBs. RFC 4122 Section 4.5 provides for cryptographically random values to be used as the node instead of the MAC address, as well as using a random value for the clock sequence when the node value changes, effectively allowing up to 61 bits of random entropy per time-slice in each identifier.

NamedGuid.Compute

Generates a deterministic name-based GUID as described in RFC 4122 Section 4.3. The following algorithms are implemented:

• MD5 generates an MD5 hash of the namespace and the name and yields a version 3 UUID.
• SHA1 generates a SHA-1 hash of the namespace and the name and yields a version 5 UUID.