Msr2020

# What constitutes Software? 
## An Empirical, Descriptive Study of Artifacts

#### MSR 2020

R.-Helge Pfeiffer, Assistant Professor, 
[Research Center for Government IT](https://www.itu.dk/forskning/institutter/institut-for-datalogi/forskningscenter-for-offentlig-it), 
[IT University of Copenhagen, Denmark](https://www.itu.dk) 
`ropf@itu.dk`

---

# What is software and why does it matter?

Is it in 2020 really necessary to research and to try to understand what software actually is? Is not the term and concept of software well understood?

I believe it is not:

- There does not seem to be a standard definition that most people refer to.

- Neither literature, industrial standards, nor standard contracts, seem to provide a consistent definition of _software_ nor a _practical_ definition of it.

---

# What is software and why does it matter?

- Stakeholders negotiating and signing software sourcing contracts are often generalists.
  - The contracts that they sign have to specify what is a deliverable and what is not. 
  - This is where the question arises: _“What shall be considered software, i.e., what precisely are the artifacts to be delivered?”_

- We know of cases where customers got delivered a systems’ source code without a complete build specification and without the built executable. Since source code alone is unusable, build code as well as executables should be considered constituents of software.

---

# What is software and why does it matter?

Since common definitions are not helpful:

> "[software are] the programs and other operating information used by a computer."
  >
  > Oxford Dictionary of English

We studied:

* Industrial Standards
  * Standard Contracts
  * Academic Articles

and we conducted an empirical experiment.

In the following some examples for these. Find more information in the paper.

---

## What do the standards say?

---

### What the standards say:

Various software engineering standards do not make completely clear what precisely software is and which artifacts constitute software.

Standards seem to agree that:
  
  - programs are fundamental to software 
  - software consists of a set of artifacts like programs, documentation, databases, configuration information, fonts, etc.

However, standards seem to disagree on which artifacts precisely constitute software. For example, are graphics and databases software or not?

---

### What the standards say:

> "computer programs, procedures, and possibly associated documentation and data pertaining to the operation of a computer system [...]"
  >
  > [_828-2012 - IEEE Standard for Configuration Management in Systems and Software Engineering_, 2012](https://ieeexplore.ieee.org/document/6170935)
  >

<hr>

> "all or a part of the programs, procedures, rules, and associated documentation of an information processing system. 
  > [...] it is typically important to include both executable and non-executable software, such as fonts, graphics, audio and video recordings, templates, dictionaries, documents and information structures such as database records."
  >
  > [_ISO/IEC 19770-5:2015(en)
 Information technology - IT asset management_, 2015](https://www.iso.org/obp/ui/#iso:std:iso-iec:19770:-5:ed-2:v1:en)

<hr>

> "all or part of the programs which process or support the processing of digital information 
  > 
  > [...] For the purposes of this definition, software excludes information per se, such as the content of documents, audio and video recordings, graphics, and databases. 
  >
  > [...] There is both executable and non-executable software. The purpose of non-executable software is to control or support executable software, and includes, for example, configuration information, fonts, and spell-checker dictionaries. Digital information which is managed by executable software (e.g. the content of documents and databases) is not considered software [...] , even though program execution may depend on data values."
  >
  > [_ISO/IEC 19770-1:2017(en) Information technology – IT asset management_, 2017](https://www.iso.org/obp/ui/#iso:std:iso-iec:19770:-1:ed-3:v1:en)

---

## What do the contracts say?

---

### Why studying standard contracts?

* In some countries public agencies use prescribed standard contracts when sourcing software. 
  * We studied the Danish standard contracts (K01, K02, and K03) and the British Model Services Contract  
  * We expect to find a crisp definition of software, the product mainly concerned by the contract.

---

### What the contracts say:

* Definitions - if given - are recursive and inconclusive.
    - MSC & K01 do not explicitly define of the term _software_.
    - K02 & K03 define it recursively.

> Software
  > 
  > Consists of Customer-specific Software and/or Standard Software. Also includes interfaces.

<hr>

> Customer-specific Software
  >
  > Software that is not classified [...] as Standard Software.

<hr>

> Standard Software
  > 
  > [...] is neither customized, adapted, nor developed specifically for the customer in connection with the supplier's fulfillment of the contract, and partly available to anyone on ordinary commercial or non-commercial and equal terms.

* Standard contracts seem to disagree on which artifacts precisely constitute software.
  * Interestingly, the lists of documentation artifacts types seem to be more precise than other potential artifacts.

---

## What do academics say?

---

### What academics say:

* Definitions - if given - are inconclusive or mere lists of exemplary artifacts:

> "software is more than just computer programs, [...] but also all associated documentation and configuration data that is required to make these programs operate correctly. [...] It may include system documentation, ... user documentation, [...] and websites for users to download recent product information."
    >
    > I. Sommerville, _Software Engineering_, 9th ed., 2010.

* Definitions are _unpractical_:

> "[...] software is pattern per se, or syntactical form [...] Hardware, in short, is also software, but only because everything is."
    >
    >  P. Suber, "What is software?" The Journal of Speculative Philosophy, pp. 89–119, 1988.

---

## What does empiricism say?

---

### What constitutes software?

> Software is the collection of any of the 19 artifacts types1 that humans collect and store and these artifacts are more than just code, documentation is an integral part of them, depending on the domain data is not necessary, and – even though rare – code is not strictly necessary.

1 source code, specific source code, script code, binary code, build code, infrastructure code, image, video, music, configuration, database, font, archive, markup, document, app data, prose, legalese, and other

How do I know?

Like a natural scientist, I executed an empirical study in which I automatically analyze more than 13 million artifacts stored in 23715 software repositories from GitHub.

---

### What constitutes software?

Each artifact is categorized into a taxonomy.

---

## Results

---

### Software is more than code

---

### Software is more than code ... but code is prominent

---

### Software is diverse

---

### Documentation an integral constituent of software

![](images/concr_cat_comb_freqs.png)

---

### Software without data?

Software without data seems to exist.

* We find 1670 repositories (ca. 7%) that do not contain any data, such as, images, videos, configuration, etc.

* On deeper investigation we find that repositories without data are most often either _libraries_, i.e., software to extend other software, _system-near software_, such as, low-level code for embedded/mobile devices, _scientific code_ that collects data at runtime, or they are just very small projects in general.

* It seems, that software in certain domains does not depend on data stored in separate artifacts.

---

### Software without code?

Quite rare but software without code seems to exits.

* Only 61 (0.26%) repositories do not contain any kind of code, i.e., neither source code, scripts, nor binary code, etc.

* No-code repositories contain either mostly prose artifacts, which is especially true for the smallest ones or they are resource storages for images, configuration, and app data.

* They are informational artifact libraries containing artifacts meant for reuse either in other software or they are training manuals/websites, or link collections.

---

### Does a characteristic distribution of frequencies of artifact categories exist?

Cannot find a certain distribution of artifact categories that is characteristic for software in general.

---

### What constitutes software?

---

### How did I generate these numbers and figures?

* To reproduce all images and numbers from the paper, use the Jupyter Notebook (Python 3) at: 
    - [NBViewer](https://nbviewer.jupyter.org/github/HelgeCPH/msr2020_what_constitutes_software/blob/master/MSR%202020.ipynb)
    - [GitHub](https://github.com/HelgeCPH/msr2020_what_constitutes_software/blob/master/MSR%202020.ipynb)
  * To replicate the entire experiment, find all code, data, and VM setup on [GitHub](https://github.com/HelgeCPH/msr2020_what_constitutes_software)

---

# What constitutes Software? 
## An Empirical, Descriptive Study of Artifacts

#### MSR 2020