# Using hostname and path conditions

> [!IMPORTANT]
> **Important:** Limiting access based on URL hostname and path can only be done with the gcloud CLI.

This page describes hostname and path [Identity and Access Management (IAM)
conditions](https://docs.cloud.google.com/iam/docs/conditions-overview).

You can secure Identity-Aware Proxy (IAP) resources using the
hostname and path of a request URL.
For example, the `request.path.startsWith` IAM condition can be
used to only grant access to employees in the **Privileged Access** group if
the URL path starts with `/admin`.

For more information about using hostname and path conditions, see
[request attributes](https://docs.cloud.google.com/iam/docs/conditions-overview#request_attributes).

### String normalization

A URL has a hostname and path. For example, the URL
`https://sheets.google.com/create?query=param`
has a hostname of `sheets.google.com` and a path of `/create`.

Backends can interpret hostnames and paths in different ways.
To remove ambiguity, IAP normalizes hostname and path
strings when checking policies.

IAP performs two policy
checks when a request has a non-normalized path. If the non-normalized
path passes the policy check, IAP normalizes the path and a
second policy check is performed. Access is granted if both the non-normalized
and normalized paths pass the policy check.

For example, if a request has the path `/internal;some_param/admin`,
IAP first performs a policy
check on the non-normalized path (`/internal`). If that check passes,
IAP performs a second policy check on the normalized path
(`/internal/admin`).

#### Hostnames

Hostname normalization involves the following:

- Removing trailing dots
- Lowercasing characters
- Converting to ASCII

Hostnames that include non-ASCII characters are further normalized using
Punycode. You need to use Punycode to normalize your hostname string for a match
to be made.

To normalize your hostname string with Punycode, use a converter
like [Punycoder](https://www.punycoder.com/).

The following are examples of normalized hostnames:

- `FOO.com` is normalized to `foo.com`
- `café.fr` is normalized to `xn--caf-dma.fr`

#### Paths

Path normalization involves the following:

- Removing path params
- Resolving relative paths to their absolute equivalent

A path param includes all
characters from a `;` to either the next `/` or the end of the path.

Requests that contain path sections starting with `..;` are considered
invalid. For example, `/..;bar/` and `/bar/..;/` return the `HTTP 400: Bad
Request` error.

The following are examples of normalized paths:

- `/internal;some_param/admin` is normalized to `/internal/admin`
- `/a/../b` is normalized to `/b`
- `/bar;param1/baz;baz;param2` is normalized to `/bar/baz`

### Subdomain endings

A policy set with `request.host.endsWith("google.com")` matches to both
`sub_domain.google.com` and `testgoogle.com`. If your goal is to limit your
policy to only subdomains that end with `google.com`, set your policy to
`request.host.endsWith(".google.com")`.