The ecoinvent database uses UUID (universal unique identifiers) to refer to many different entities: exchanges, datasets, geographies, etc. Following it is explained the use of those UUID and which UUID changes or remains constant across versions.
An undefined dataset (before linking) is identified by an id number (ecospold2 format field #102). Associated with this number is a unique combination of activityName, geography and time period. For example, the UUID 013b2042-2ed2-4534-a35e-272e0deb4b34 refers to dioxane production, RER, valid from 2003-01-01 to 2016-12-31.
The activityName and the geography also have their UUID. The UUID a99b42d5-29a0-4543-87cf-44fec5aaa76d refers to "dioxane production", and 0723d252-7e2a-11de-9820-0019e336be3a refers to "RER". Time periods do not have UUIDs.
From one version of ecoinvent to the other, undefined datasets do not change their UUID, if their combination of activityName UUID, geography UUID and time period has remained constant. This is the case of dioxane production.
Sometimes, an activityName has been changed, but still refers to the same UUID. This is the case of 760e60db-b6d2-4964-826c-f096f3a6b2f8. In version 3.2, the name is "rock wool production", and in 3.3, "stone wool production". Because the UUID is the same, even if the name has changed, the UUID bbd22617-d702-45b9-8af7-d102ae7cab4c is found in both 3.2 and 3.3, referring to either "rock wool production, CH, 2000-01-01 to 2007-12-31" or "stone wool production, CH, 2000-01-01 to 2007-12-31".
If a dataset's time period is changed from one version to the other, the UUID is also changed. Ecoinvent does not support time series yet, so if a time period is changed, only the most recent dataset is part of the distributed database.
If some data is changed inside the dataset, for example values of exchanges, comments, etc., the UUID is not changed.
After linking, the UUID of a dataset is changed. UUIDs for the same dataset in different system models are different.
After linking, the same combination of activityName UUID, geography UUID and time period can occur many times, because multi-output datasets might have resulted in many single-output datasets.
The same combination of system model, activityName UUID, geography UUID, time period and reference product, for different versions of ecoinvent, yields different UUID. For example, the cut-off version of dioxane production, RER, 2003-01-01 to 2016-12-31, has the UUID f9fcfe4c-31e0-4118-85fc-ad64b856fd62 in 3.2 and c8744c85-2876-4f68-9e02-bf0a920d4804 in 3.3.
Exchange names have UUID that are consistent across versions and datasets. The field referring to this information is called "elementaryExchangeId" or "intermediateExchangeId". This means that for example, "formaldehyde", has the same UUID wherever it appears.
Exchanges also have an id number, different each time it appears within the same version of the database. If two datasets use "formaldehyde" from technosphere, the intermediateExchangeId will be identical, but the id number will be different. Also, the id number of an exchange is different before and after linking, and across different system model. So, the id of an exchange is kept the same in the undefined datasets across versions but the id of an exchange is different for the same linked dataset across versions.
The same intermediate exchange can appear many times in the same dataset, with different UUID, because it can be for example the reference product and from technosphere (in a market or a market group), or because it is supplied by different activities through activity links.