[http, https | telnet, tn3270, rlogin | gopher | file | ftp | wais | news, nntp, snews | newspost, newsreply, snewspost, snewsreply | mailto | finger | cso | bibp | lynxexec, lynxprog | lynxcgi| NcFTP | internal]

URL Schemes Supported in Lynx

Lynx handles a number of URL types, that are enumerated below. For more details about URLs (Uniform Resource Locators) see RFC1738:

Lynx resolves partial or relative URLs in documents with respect to the BASE if one was specified, otherwise with respect to the document's absolute URL, using the rules described in RFC1808:

and in subsequent drafts of the IETF:

When entering a URL on the command line to be used as the startfile, or at the prompt for a “g”oto entry, a partial host field can be used and the scheme field can be omitted if the scheme and fully qualified domain name can be constructed internally by using the URL_DOMAIN_PREFIXES and URL_DOMAIN_SUFFIXES definitions in the Lynx configuration file. See the explanation of those definitions and their use in your lynx.cfg.

For example, wfbr will be treated as http://www.wfbr.edu/, and wfbr/dir/lynx will be treated as http://www.wfbr.edu/dir/lynx, but gopher.wfbr.edu/11/_fileserv/_lynx will be treated as gopher://gopher.wfbr.edu/11/_fileserv/_lynx.

For files or directories on the local host, a tilde (~) is expanded to the path of the account's login directory, e.g., ~/foo will be expanded to file://localhost/your/login/directory/foo. The tilde expansion is done homologously on Unix and VMS.

On VMS, Lynx also will expand any file or directory spec recognizable to DCL into a valid URL, e.g., [] will be expanded to file://localhost/current/default/directory.

These expansions are SOLELY for startfile or “g”oto entries! Any partial or relative URLs within HTML documents are resolved according to the rules specified in RFC1808 and subsequent IETF drafts.


The http and https URLs:

Lynx handles http URLs exactly as specified in RFC1738. The format is:

      http://host:port/path?searchpart#fragment

where :port is optional and defaults to :80, /path if present is a slash-separated series of symbolic elements, and ?searchpart if present is the query for an ISINDEX search or the content of a FORM with METHOD="GET". The #fragment field if present indicates a location in the document to seek for display, based on a NAME-ed anchor or an ID attribute within the document, and is technically an instruction rather than part of the URL. Lynx will treat ID attributes as NAME-ed anchors for all tags in the BODY of a document which can correspond to positions in the rendering of the document.

The https URL has the same format, but the default port is :443.

Lynx relies for https support on external libraries (OpenSSL or GnuTLS) whose capabilities have evolved over time. In turn, those libraries may depend upon external resources for verifying SSL certificates. For instance, certification revocation may be provided via the Online Certificate Status Protocol (OCSP) which is an external service. Without this facility, Lynx may not warn about websites using revoked SSL certificates.


The telnet, tn3270, and rlogin URLs:

A telnet URL generally results in Lynx spawning a telnet session. Lynx implements the complete telnet URL scheme, i.e.:

      telnet://user:password@host:port

The user and/or :password fields may be omitted, and the @ should be omitted if neither is present. The port defaults to :23 when omitted in the URL.

A tn3270 or rlogin URL is specified equivalently, and similarly spawns a tn3270 or rlogin session. The actual behavior is dependent on the TCP-IP software installed on the local and target hosts.

It is unwise to include the :password field except for URLs which point to anonymous or other public access accounts, and for most TCP-IP software you will be prompted for a password whether or not one was included in the URL.


The gopher URL:

The gopher URL takes the form:

      gopher://host:port/gopher-path

where :port is optional and defaults to :70, and the /gopher-path is opaque (not fully equivalent to the slash-separated series of symbolic elements of http paths) as explained in RFC1738. Typically, the gopher-path consists of a gophertype indicating the file or service type (e.g., 0 or I for plain text or an image, respectively, 7 for a search, or 1 for a directory), followed by a platform-specific selector. Any reserved characters in the selector should be hex escaped (%hh), including slashes, although hex escaping of slashes is not required by Lynx in gopher URLs.

Lynx does not overtly support the gopher+ protocol, and does not represent itself as gopher+ capable when communicating with gopher servers. Lynx might transmit any (hex-escaped-tab-separated) extended gopher+ fields in a URL if an author included them in a document, but is likely to mishandle what the gopher server returns in such cases, and would not generate and transmit them itself. For pre-formed URLs to submit gopher searches, it may be better to use a ? rather than hex-escaped tab (%09) as the separator for the searchpart in the selector, e.g.:
gopher://gopher.wfbr.edu/77/_shell/search.shell%20/_shell/walker?lynx* Lynx will handle the %09 if you use that instead of ?, but other WWW clients may mishandle it.

For the gophertype which signifies HTML (h), if the selector begins with GET%20/ Lynx will convert the gopher URL to an http URL, e.g.:

gopher://www.wfbr.edu:80/hGET%20/

will become:

http://www.wfbr.edu/

The port field will be retained if it is not :80, and will default to :70 if it was defaulted originally. These conventions were adopted during development of the University of Minnesota gopher software to facilitate the offering of links to MIME-capable http servers in the listings returned by gopher servers, but should be considered Lynxisms and UMN Gopherisms.


The file URL:

The file URL is used to retrieve files or generate a directory listing on the local host. The host field can be localhost or a domain name for the local host:

file://localhost/path

If you do not use localhost or a domain name for the local host, Lynx will substitute ftp:// for file:// and treat it as an ftp URL.

The /path is treated as originating at the root, unless you include a tilde (~), e.g.:

      file://localhost/~/foo   will be converted to:
      file://localhost/your/login/directory/foo

The latter feature is a Lynxism, is done homologously on Unix and VMS, and should be used ONLY in local documents intended for Lynx.

On VMS, the first element of the path, if not a tilde, is assumed to be a device, e.g.:

      file://localhost/www_root/directory/filename.suffix

should be used for: www_root:[directory]filename.suffix
If you are unsure how to specify a file URL in local documents on VMS, invoke Lynx with the desired file or directory as the startfile using any spec acceptable to DCL, and then use the showinfo command (=) to see the file URL which Lynx created for it.


The ftp URL:

The ftp URL has the general format:

      ftp://host:port/path;type=[D,I, or A]
      ftp://username@host:port/path;type=[D,I, or A]

The default port is :21 and the default username is anonymous. If username is included, Lynx will prompt you for the password. For anonymous ftp, Lynx uses your personal_mail_address (user@host) as the password if it has been defined via the “o”ptions menu. Otherwise, Lynx uses the dummy password WWWUser. (A password can also be embedded in the URL, by replacing username with username:password. This is strongly discouraged for “real” passwords that must be kept secret, since URLs with the completely unencrypted password may show up on the screen, in HISTORY and LIST pages etc., and may even become visible to remote sites for example through Referer headers.) Do not include the @ if neither username nor :password is included.

The ;type= parameter can be used with value D, I, or A to force handling of the URL as, respectively, a directory listing, binary file, or ASCII file. The Lynx ftp gateway normally determines this itself, but the parameter can be used if the internal procedure draws an incorrect inference about the nature of the ftp URL.

The /path is treated according to RFC1738 for VMS and VM/CMS ftp servers. The lead slash (/) is treated purely as a separator, not as a designator for the root, and the path string if present is treated as in or under the login directory. For VMS ftp servers, if you wish to have the first element treated as a device rather than file or subdirectory name, begin it with a hex-escaped slash (%2f), e.g.:

      ftp://user@myhost/%2fsys$common/syshlp

can be used for a listing of sys$common:[syshlp]
Also, on VM/CMS ftp servers, if the path string begins with vmsysu%3a it receives special handling as an SFS path, e.g.:

      ftp://ubvm.cc.buffalo.edu/vmsysu%3alistserv.webshare

For Unix and Unix-emulation ftp servers, RFC1738 is not respected and the lead slash is treated as the root, i.e., the /path is handled equivalently to that in file URLs. The distinction is irrelevant for anonymous ftp, but matters when using ftp for non-anonymous accounts. If you are using ftp with a Unix server and do wish to get a listing of the login directory or have the path string treated as a file or path under the login directory, include a tilde (~) as for file URLs, e.g.:

      ftp://user@myhost/~

The wais URL:

The wais URL is used to retrieve resources using the Wide Area Information System protocol. The format is:

      wais://host:port/database
      wais://host:port/database?wais_query
      wais://host:port/database/wais_type/wais_path

where :port defaults to :210

Direct wais support is built into Lynx for VMS, and can be compiled into Lynx on Unix.

If only a database is indicated in the URL, Lynx returns an ISINDEX cover page for searching that database, and will submit your search with the wais_query appended. Lynx will convert the server's reply into a hit list with URLs that include the wais_type and wais_path for retrieving items from the hit list.


The news, nntp, and snews URLs:

The news and nntp URLs are handled by Lynx as specified in RFC1738, but for compatibility with other clients, Lynx allows inclusion of host and port fields in news URLs, which properly should be used only in nntp and snews URLs. If not included in news URLs, Lynx will use the nntp server pointed to by the NNTPSERVER environment variable or configuration symbol (see lynx.cfg), with default port :119. A host field must be included in nntp URLs, and the port field is optional with the same default.

If the URL requires authentication, Lynx will prompt you for the username and password. These are cached during a session, for reuse on the same host. If $HOME/.newsauth exists, Lynx initializes its cache from this file. The .newsauth file contents are one line per entry: hostname, password and username (in that order) separated by a space.

The formats are:

      news:newsgroup (retrieves list of messages in newsgroup)
      news:messageID (retrieves the message)
      news:* (retrieves list of all available newsgroups)
      nntp://host:port/newsgroup
      nntp://host:port/messageID
      nntp://host:port/*

(snews same as nntp, but the default port is :563)

The messageID is the message's unique identifier, consisting of an identification string and the host of origin for the message (ident_string@origin_host).

Lynx also supports wildcarding via an asterisk for listings of news hierarchies or sub-hierarchies, e.g.:

      news:comp.infosystems.*
      nntp://host:port/comp.infosystems.*

(snews same as nntp, but the default port is :563)
This is not in RFC1738 and may not be supported by all other clients.

Lynx allows you both to reply to the author of a news message via email, and, if news posting has been enabled, to send a followup message to the newsgroup (see newspost, newsreply, snewspost, snewsreply).

Lynx converts any strings in news messages which appear to be a URL with a supported scheme into a link for accessing that URL.

Lynx also supports the newsgroup and message number URL scheme:

      news:newsgroup/startNo-endNo (lists message range in newsgroup)
      news:newsgroup/messageNo     (retrieves the message by number)
      nntp://host:port/newsgroup/startNo-endNo
      nntp://host:port/newsgroup/messageNo

(snews same as nntp, but the default port is :563)
Use of this scheme is not recommended, because the message numbers are specific to each nntp server, unlike the unique identifiers for news messages.


The newspost, newsreply, snewspost, and snewsreply URLs:

When Lynx receives group listings or articles via news, nntp or snews URLs, it also checks whether the nntp server supports posting from the Lynx user's site, and if so, includes links for posting new messages to that server, or for posting followups (replies) to previously posted messages. RFC1738, and IETF URL drafts through this release of Lynx, do not include any schemes for posting to news groups. Lynx has long supported newspost and newreply URL schemes for posting new messages or sending followups, respectively, to standard nntp servers, with default port :119. Lynx now also supports homologous snewspost and snewsreply URLs for use with SSL capable nntp servers.

The formats are:

      newspost://host:port/newsgroup(s)  (post a new message)
      newsreply://host:port/newsgroup(s) (post a followup message)

(snewspost and snewsreply have the same formats, but the default port is :563)

If the host field is omitted, it defaults to that pointed to by the NNTPSERVER configuration or environmental variable. Inclusion of at least one newsgroup in the URL is required, and additional groups can be specified as a comma-separated list. Wildcarding of newsgroup names is not supported for these URLs. For newsreply and snewsreply URLs, if an external editor has been defined via the Options Menu, the user is offered an option to include the currently displayed document, which presumably is a news article with a followup link that was activated, and if confirmed, each line of that document is prefixed with a right-angle-bracket. The user is expected to edit such an inclusion so that only the passages relevant to the followup message are retained.

These URLs can be used as command line startfiles (in which case, Lynx will exit after posting the message, and the newreply or snewsreply URLs degrade to newspost or snewpost URLs, respectively). They also can be used as HREF attribute values in any HTML document homologously to mailto URLs, with the qualification that they presently are supported only by Lynx.


The mailto URL:

The mailto URL is used to provide links that when activated can be used to send a comment or the content of a FORM to an Internet email address (user@host). The format is:

      mailto:user@host

The description of the mailto URL in RFC1738 has been interpreted by some as allowing only a single recipient, but Lynx invented the mailto URL, has always supported a series of user@host addresses as a comma-separated list, and still does. For compatibility with Explorer, Lynx also accepts a semi-colon-separated list.

For compatibility with Netscape, Lynx parses any ?subject=The%20Subject appended to the URL, trims the URL at the ?, and uses the value as the default Subject: for the message or FORM content mailing. This is not recommended practice. The preferred way to indicate the default Subject: for a LINK or Anchor with a mailto HREF, or a FORM with a mailto ACTION, is via a TITLE attribute with the subject string as its value, e.g.:

      <LINK REV="made"
            HREF="mailto:me@myhost,her@herhost" TITLE="The Subject">

      <A HREF="mailto:user@host" TITLE="The Subject">...</A>

      <FORM METHOD="post" ENCTYPE="text/plain"
            ACTION="mailto:WebMaster@host" TITLE="The Subject">
       ...
      </FORM>

Note that a TITLE attribute for FORM is now included in the HTML specifications. Some clients use a SUBJECT attribute for this purpose in FORM tags, and Lynx recognizes that as a synonym for TITLE.

Lynx also will process any to=address(es), cc=address(es), keywords=word_list and/or body=message fields in ?searchpart tack-ons to mailto URLs. The to and/or cc values can be single addresses, or comma- or semi-colon-separated lists of addresses. All addresses, and any body values, will be offered for approval by the user before proceeding with a mailing. Any other name=value pairs in the ?searchpart will be ignored. Also, if the mailto URL is the ACTION for a FORM, any body in a ?searchpart tack-on will be ignored, because the body of the mailing must be constructed solely from the the FORM's content. Lynx expects multiple name=value pairs in a ?searchpart tack-on to be separated by ampersands, as in the original Netscape implementation, and in an equally ill-advised IETF draft of that implementation (draft-hoffman-mailto-url-03.txt). These should be represented as entities (&amp;) in the HTML markup. This functionality is generally desired, but the IETF backward compatibility principal normally would lead to a new scheme being used (e.g., mail:, or smtp:), rather than breaking mailto: implementations.

If ENCTYPE="text/plain" is specified for a FORM with a mailto ACTION, Lynx will not hex escape the name=value pairs of the FORM's content, and will use physical newlines instead of “&” or “;” to separate the pairs, so that the content will be readable directly. Otherwise, Lynx will mail the content with the default:

      ENCTYPE="application/x-www-form-urlencoded" (“&” separates pairs)

or:

      ENCTYPE="application/sgml-form-urlencoded" (“;” separates pairs)

if the latter was indicated.

Note that when mailing FORM content Lynx wraps any lines longer than 78 characters, to avoid buffer overflows in mail software and to ensure reliable transmission across gateways. If the ENCTYPE was not text/plain, any script which decodes the mailed content should ignore the physical newlines and recognize only hex escaped newline characters as intended to be present in the decoded content.

If the mailto URL is not the ACTION for a FORM, and if an external editor has been defined via the Options Menu, the user is offered an option to include the currently displayed document. If this option is accepted, each line of that document is prefixed with a right-angle-bracket, and the prefixed inclusion should be trimmed by the user to just those passages relevant to the message which will be sent.


The finger URL:

Lynx has full support for the finger protocol, but a format for finger URLs has not yet been adopted by the IETF. The formats supported by Lynx therefore include every possibility not inconsistent with RFC1738, including:

      finger://host                         finger://@host
      finger://host/                        finger://@host/
      finger://host/%2fw                    finger://@host/w
      finger://host/w                       finger://host/w/
      finger://host/username[@host]         finger://username@host
      finger://host/username[@host]/        finger://username@host/
      finger://host/w/username[@host]       finger://username@host/w
      finger://host/%2fw%20username[@host]  finger://host/username[@host]/w
      finger://host/w/username

Activating a finger URL will send a request to the finger server via port 79 on the host specified. You can include :79 in the URL, but no other value is allowed. The /w or /%2fw is used to request a full report for finger servers which support it, and is not case sensitive (i.e., can be /W or /%2fW). Any strings in the report which appear to be a URL with a supported scheme will be converted into a link for accessing that URL.

An alternative way to access finger servers is via gopher URLs with port 79 and the plain text (0) gophertype specified:
gopher://host:79/0
Lynx will handle such URLs equivalently to overt finger URLs, including creation of links for any strings which appear to be supported URLs.


The cso URL:

The cso URL is intended to provide a gateway to CSO/PH (QI) servers. The requests are made on port 105 by default (:105), with the following overt cso URL format:

      cso://host

You also can use a gopher URL format with port 105 and the CSO (2) gophertype specified:

      gopher://host:105/2

Lynx will parse the stream returned by the server for the above URLs and create a FORM for submitting additional requests (searches) to the server. Any strings in the reports returned for these requests (searches) which appear to be a URL with a supported scheme will be converted into a link for accessing that URL.


The bibp URL:

Lynx provides built-in support for bibliographic protocol (BibP). BibP links are links to published works such as books or journal articles, without a predefined server. BibP links are intended for resolution by a local bibhost server (http://bibhost/) if it exists. Otherwise, resolution is performed by a document-specified server or a known global server.

The lynxexec and lynxprog URLs:

If execution of spawned commands has been enabled in your Lynx image, the lynxexec and lynxprog URLs can be used to execute arbitrary system commands or invoke system utilities. Any system command and associated switches or qualifiers can be used, with the syntax appropriate for a shell running Lynx on Unix, or for DCL on VMS, e.g.:

      lynxexec:dir/date/size foo:[blah] (VMS)
      lynxexec:ls -l /foo/blah          (Unix)
      lynxprog:news

(Note, however, that restrictions on acceptable commands or utilities may be imposed by the system administrator.)

You optionally can include //localhost/ in the URL, between the scheme field and the command, but that is always implied. The lynxexec and lynxprog URLs differ only in that with lynxexec you are prompted to enter RETURN before Lynx clears the screen and restores the previously displayed document, so that you can read any screen output generated by the spawned command, whereas no such pause is imposed upon exit from the utility invoked via lynxprog.

These are Lynxisms and should be used only in local documents intended solely for Lynx.


The lynxcgi URL:

The lynxcgi URL is implemented only on Unix, can be used as the ACTION for a FORM, and if enabled in your Lynx image has the format:

      lynxcgi://localhost/path_to_CGI_script

where //localhost is optional and always implied; the full path should be specified, as “~” is not recognized; if the script is in the directory Lynx was started from, the simple file name is adequate. The output of the script should be text/html and is rendered and displayed by Lynx. Restrictions on use of lynxcgi and on acceptable paths can be imposed in userdefs.h and lynx.cfg, qv.

This is a Lynxism and should be used only in local documents intended solely for Lynx, or for limited local testing of CGI scripts without an http server.


The NcFTP URL:

Lynx recognizes the NcFTP-style ftp URL, e.g.,

        ftpHost:fileSpecification

for example


        ftp.gnu.org:/pub/gnu


The LYNXfoo internal URLs:

Lynx uses a variety of private URL schemes for communication among its internal modules. They start with uppercase letters LYNX by convention, although, as input, URL schemes are recognized in a case-insensitive manner.

As you discover what they are, and are tempted to use them externally in documents, you should resist that temptation:

For example, tempting though it might be, do not use these:

      Return to your <A HREF="LYNXHIST:0">Startfile</A>
      Review your <A HREF="LYNXKEYMAP:">Keymap</A>

(No, they will not do any harm. Yes, they work. But do not rely on it.)

If you must try one, the second is OK from the command line:

      lynx LYNXKEYMAP:

But within Lynx, use the “K” keystroke command. Sometimes it may be convenient to use a private scheme with “g”oto, as in:

      g LYNXMESSAGES:
      g LYNXCOMPILEOPTS:
      g LYNXCFG:

But again, there usually is a way in which those special pages are meant to be reached that is more convenient.