"When defining a surrogate primary key for a [SQL Server] table, two options are the most common: Integer and UniqueIdentifier (aka Globally Unique Identifiers, or GUID's) ... Historically, Integer has been the logical choice. It’s human-readable, requires minimal storage, and can be set as an identity (auto-incrementing) to prevent the need for additional application logic. UniqueIdentifier comes with significant disadvantages. The most immediately noticeable is that it’s user-unfriendly. You’ll never hear a user or developer ask you about record “A78383A3-4AB1-42CF-B3FC-A4A23AD10398”. With high availability and replication becoming highly prevalent, UniqueIdentifier is being chosen more often, but has caveats that mean it isn’t always the optimal solution."

--Jeffrey J. Keller, Vertabelo.com

"Many data and information modelers talk about all kinds of keys (or identifiers. I'll forego the distinction for now). I hear them talk about primary keys, alternate keys, surrogate keys, technical keys, functional keys, intelligent keys, business keys (for a Data Vault), human keys, natural keys, artificial keys, composite keys, warehouse keys or Dimensional Keys (or Data Warehousing) and whatnot. Then a debate rises on the use (and misuse) of all these keys ... The foremost question we should actually ask ourselves: can we formally disambiguate kinds of keys (at all)? Of all kinds of key, the primary key and the surrogate key gained the most discussion."

"If we take a look at the relational model we only see of one or more attributes that are unique for each tuple in a relation -- no other formal distinction is possible. When we talk about different kinds of keys we base our nomenclature on properties and behavior of the candidate keys. We formally do not have a primary key, it is a choice we make and as such we might treat this key slightly different from all other available keys in a relation. The discussion around primary keys stems more from SQL NULL problems, foreign key constraints and implementing surrogate keys."

"The Internet is full of dogmatic commandments for choosing and using keys in relational databases. At times it verges on a holy war: should you use natural or artificial keys? Auto-incrementing integers, UUIDs? After wading through sixty-four articles, skimming sections in five books, and asking questions on IRC and StackOverflow I think I’ve put the pieces together and have a recommendation to harmonize the various camps. Many arguments about keys boil down to false dichotomies and failures to acknowledge other points of view."

--Joe Nelson, begriffs.com

"I'm seeing more [data] professionals implementing the following type of conditional unique constraints, typically related to the use of 'soft deletes'):

TABLEX (SK,A,B,C,ACT_FLAG)

Uniqueness is defined for (A,B) iff ACT_FLAG='Yes'. Any row with a ACT_FLAG = 'No' is excluded from the unique requirement."

"Most SQL DB implementations I know of do not allow this type of constraint to be enforced declaratively. Instead it relies on tricks within the index specification for enforcement."

"This conditional application of unique constraints troubles me. The prevalent use of a surrogate primary key avoids duplicates in the table as a whole. But this approach seems to declare a business rule that can be turned off and on based on the value of a non-key column. It definitely feels there is more than one kind of business entity here -- but they do share the same attributes. I also see a similar design pattern when OO classes are mapped to a table during implementation (when the implementation approach is to combine classes into a single table and an attempt is made to enforce two types of 'uniqueness'."

"However, I cannot definitively find a specific rule/guideline within relational theory that it violates. I know there are design alternatives. The fact that there is no true always-on business key other than the surrogate key IS an issue, but the fact that technically the surrogate PK prevents duplicates is almost always presented as a counter argument."

"To the question “How relevant is data modeling in the world of NoSQL?” I give the following answer.

The main purpose of data modeling is to understand the business, some application domain, some users world. The model becomes a representation of that world -- the "things" in it, the relationships among those things and any constraints on those things or relationships. A secondary purpose is to build a database to contain information which pertains to and describes that domain."

"Generally we speak of the model coming first, then the implementation, and finally, the data gets collected and stored according to the model. Hence, the business data model should not be concerned with issues of physical stored representation, or the transformations/manipulations/constraints which are imposed to facilitate implementation in some data (storage) management system. That could be a relational DBMS, or a NoSQL tool".

" ... increasingly the data already exists in some form. Which leaves us with the task of figuring out what it means, what it represents -- that is, understanding the data as it represents some user domain. NoSQL tools are often designed to deal with existing data and to process it more efficiently (that may be an oversimplification!). Either way, you must understand the business in order to make sense of the data." 

--Gordon Everest, LinkedIn.com