[c#] Guid.NewGuid() vs. new Guid()

What's the difference between Guid.NewGuid() and new Guid()?

Which one is preferred?

Guid.NewGuid(), as it creates GUIDs as intended.

Guid.NewGuid() creates an empty Guid object, initializes it by calling CoCreateGuid and returns the object.

new Guid() merely creates an empty GUID (all zeros, I think).

I guess they had to make the constructor public as Guid is a struct.

[I understand this is an old thread, just adding some more detail] The two answers by Mark and Jon Hanna sum up the differences, albeit it may interest some that

Guid.NewGuid()

Eventually calls CoCreateGuid (a COM call to Ole32) (reference here) and the actual work is done by UuidCreate.

Guid.Empty is meant to be used to check if a Guid contains all zeroes. This could also be done via comparing the value of the Guid in question with new Guid()

So, if you need a unique identifier, the answer is Guid.NewGuid()

Guid.NewGuid() creates a new UUID using an algorithm that is designed to make collisions very, very unlikely.

new Guid() creates a UUID that is all-zeros.

Generally you would prefer the former, because that's the point of a UUID (unless you're receiving it from somewhere else of course).

There are cases where you do indeed want an all-zero UUID, but in this case Guid.Empty or default(Guid) is clearer about your intent, and there's less chance of someone reading it expecting a unique value had been created.

In all, new Guid() isn't that useful due to this lack of clarity, but it's not possible to have a value-type that doesn't have a parameterless constructor that returns an all-zeros-and-nulls value.

Edit: Actually, it is possible to have a parameterless constructor on a value type that doesn't set everything to zero and null, but you can't do it in C#, and the rules about when it will be called and when there will just be an all-zero struct created are confusing, so it's not a good idea anyway.