4.3a Read, Write, Execute bits for User, Group, Other — decoding ls -la output¶

🧠 Context Introduction¶

When working with AI infrastructure, you will frequently need to manage access to critical files — model weights, training datasets, configuration files, and scripts. Linux permissions determine who can read, write, or execute these resources. The command ls -la is your primary tool for inspecting these permissions. This section breaks down how to decode that output, so you can confidently control access to AI assets.

⚙️ The Anatomy of a Permission String¶

Every file and directory in Linux has a permission string displayed in the first column of ls -la output. It looks like this:

-rwxr-xr--

This string is 10 characters long and is divided into four parts:

🕵️ Decoding the Three Permission Bits¶

Each set of three characters (positions 2-4, 5-7, 8-10) uses the same pattern:

• r = Read permission (view file contents or list directory)
• w = Write permission (modify file or create/delete files in directory)
• x = Execute permission (run a script or enter a directory)
• - = Permission is denied (no access for that bit)

For example, rwx means full access: read, write, and execute. r-- means read-only. --- means no access at all.

📊 Permission Breakdown Table¶

Here is how to interpret the full permission string -rwxr-xr--:

Each permission set can also be represented as a numeric (octal) value: - r = 4 - w = 2 - x = 1 - - = 0

Add the values together for each segment. rwx = 4+2+1 = 7. r-x = 4+0+1 = 5. r-- = 4+0+0 = 4.

🛠️ Reading a Full ls -la Line¶

When you run ls -la, you see output like this (formatted as bolded inline text):

-rwxr-xr-- 1 alice ml-team 2048 Mar 15 09:30 train_model.py

Breaking this down:

• -rwxr-xr-- — Permission string (file type + user + group + other)
• 1 — Number of hard links
• alice — File owner (User)
• ml-team — Group owner
• 2048 — File size in bytes
• Mar 15 09:30 — Last modification timestamp
• train_model.py — File name

The key takeaway: alice (the user) can read, write, and execute this script. Anyone in the ml-team group can read and execute it. Everyone else can only read it.

📊 Visual Representation: Linux Permission String Anatomy¶

This diagram maps out a standard 10-character file permission string (-rwxr-xr--), breaking it down by entity class (User, Group, Other) and individual permission bits with their numeric weight.

🔐 Common Permission Scenarios for AI Infrastructure¶

Here are typical permission patterns you will encounter:

• Model weights file — -rw------- (Owner only can read/write; protects sensitive model data)
• Training script — -rwxr-xr-x (Owner can modify; group and others can execute)
• Dataset directory — drwxr-x--- (Owner full access; group can read/list; others denied)
• Configuration file — -rw-r--r-- (Owner can edit; everyone else can read)
• Log file — -rw-rw---- (Owner and group can read/write; others denied)

🧪 Quick Reference: What Each Permission Allows¶

✅ Summary Checklist for Decoding ls -la¶

• ✅ First character tells you if it is a file (-) or directory (d)
• ✅ Next three characters = User (owner) permissions
• ✅ Next three characters = Group permissions
• ✅ Last three characters = Other (everyone else) permissions
• ✅ A dash (-) means that specific permission is denied
• ✅ Use the numeric values (4, 2, 1) to quickly calculate permission sets
• ✅ Always check permissions before running AI scripts or accessing datasets to avoid errors

Understanding these bits is essential for securing AI infrastructure — you control exactly who can read training data, who can modify model code, and who can execute inference scripts.