842

When one wants to refer to some part of a webpage with the "http://example.com/#foo" method, should one use

<h1><a name="foo"/>Foo Title</h1>

or

<h1 id="foo">Foo Title</h1>

They both work, but are they equal, or do they have semantic differences?

mikemaccana
  • 110,530
  • 99
  • 389
  • 494
Henrik Paul
  • 66,919
  • 31
  • 85
  • 96
  • 2
    The link should actually be `http://example.com#foo` (so without the / before #) – Dana Sep 10 '13 at 19:53
  • 87
    Actually, `http://example.com#foo` and `http://example.com/#foo` are equivalent as defined in one of the RFCs on URIs. – Oliver Oct 29 '13 at 08:29
  • 1
    Please, check this answer: https://stackoverflow.com/a/69869066/2457251 – Almir Campos Nov 07 '21 at 01:25
  • Note that `

    ` can be referenced from `window.foo` in Javascript, which may not be what you are expecting, see https://stackoverflow.com/a/28568419/247696
    – Flimm Oct 06 '22 at 10:00

15 Answers15

661

According to the HTML 5 specification, 5.9.8 Navigating to a fragment identifier:

For HTML documents (and the text/html MIME type), the following processing model must be followed to determine what the indicated part of the document is.

  1. Parse the URL, and let fragid be the <fragment> component of the URL.
  2. If fragid is the empty string, then the indicated part of the document is the top of the document.
  3. If there is an element in the DOM that has an ID exactly equal to fragid, then the first such element in tree order is the indicated part of the document; stop the algorithm here.
  4. If there is an a element in the DOM that has a name attribute whose value is exactly equal to fragid, then the first such element in tree order is the indicated part of the document; stop the algorithm here.
  5. Otherwise, there is no indicated part of the document.

So, it will look for id="foo", and then will follow to name="foo"

Edit: As pointed out by @hsivonen, in HTML5 the a element has no name attribute. However, the above rules still apply to other named elements.

Dan Rosenstark
  • 68,471
  • 58
  • 283
  • 421
Greg
  • 316,276
  • 54
  • 369
  • 333
196

You shouldn’t use <h1><a name="foo"/>Foo Title</h1> in any flavor of HTML served as text/html, because the XML empty element syntax isn’t supported in text/html. However, <h1><a name="foo">Foo Title</a></h1> is OK in HTML4. It is not valid in HTML5 as currently drafted.

<h1 id="foo">Foo Title</h1> is OK in both HTML4 and HTML5. This won’t work in Netscape 4, but you’ll probably use a dozen other features that don’t work in Netscape 4.

hsivonen
  • 7,908
  • 1
  • 30
  • 35
  • 7
    +1 for talking about browser support. Is NS4 the only one not supporting url#id => element.id? – Hashbrown Oct 22 '13 at 00:06
  • 17
    @Hashbrown Couldn't find an answer, so I did some testing. I found that even very old browsers treat `id`s just like `name` anchors in terms of URL fragments and [compatibility](https://developer.mozilla.org/en-US/docs/Web/CSS/:target#Browser_compatibility) of the CSS `:target` selector. Tested: Chrome 6, Firefox 1.5, IE6, Opera 8.02, Safari 3.1.2, Netscape 7.2, Lynx 2.24, and mobile browsers: Android 2.2, Chrome 26, Dolphin 9.3, Firefox 19, IE10, Safari 4, and Opera Mini 5.1. – Stephen M. Harris Feb 07 '14 at 19:01
  • 1
    @smhmic, I found one. The Off-By-One web browser recognizes anchors defined via but won't recognize anchors defined via OB1 was last updated >8 years ago. Its author boasts that it "may be the world's smallest and fastest web browser with full HTML 3.2 support." It claims support for Win95 thru XP, but it works fine w/ 64-bit Win7. So, why have such a dinosaur? For testing, of course, to make sure my sites don't break too badly w/ truly antique browsers. Also, I also carry OB1 on a flash drive. It is tiny, self-contained & immune from infections. – Dave Burton Oct 04 '14 at 14:30
  • 37
    Reading this in 2016 be like.. *Netscape 4?* – ADTC Aug 14 '16 at 13:52
  • Im using Outlook the enterprise email app i guess and when I put `scroll to element` it doesnt work, but using `name` instead of `id` works which is sad. – Coty Embry Mar 01 '18 at 16:39
  • 1
56

I have to say if you are going to be linking to that area in the page... such as page.html#foo and Foo Title isn't a link you should be using:

<h1 id="foo">Foo Title</h1>

If you instead put an <a> reference around it your headline will be influenced by an <a> specific CSS within your site. It's just extra markup, and you shouldn't need it. I'd highly recommend placing an id on the headline, not only is it better formed, but it will allow you to either address that object in Javascript or CSS.

Zano
  • 2,595
  • 27
  • 33
Tim Knight
  • 8,346
  • 4
  • 33
  • 32
30

Wikipedia makes heavy use of this feature like this:

<a href="#History">[...]</a>
<span class="mw-headline" id="History">History</span>

And Wikipedia is working for everybody, so I would feel safe sticking with this form.

Also don't forget, you can use this not only with spans but with divs or even table cells, and then you have access to the :target pseudo-class on the element. Just watch out not to change the width, like with bold text, cause that moves content around, which is disturbing.

Named anchors - my vote is to avoid:

  • "Names and ids are in the same namespace..." - Two attributes with the same namespace is just crazy. Let's just say deprecated already.
  • "Anchors elements without href atribute" - Yet again, the nature of an element (hyperlink or not) is defined by having an atribute?! Double crazy. Common sense says to avoid it altogether.
  • If you ever style an anchor without a pseudo-class, the styling applies to each. In CSS3 you can get around this with attribute selectors (or same styling for each pseudoclass), but still it's a workaround. This usually doesn't come up because you choose colors per pseudo-class, and the underline being present by default it only makes sense to remove, which makes it the same as other text. But you ever decide to make your links bold, it'll cause trouble.
  • Netscape 4 might not support the id feature, but still an unknown attribute won't cause any trouble. That's what called compatibility for me.
Zoltán Morvai
  • 521
  • 4
  • 7
15
<h1 id="foo">Foo Title</h1>

is what should be used. Don't use an anchor unless you want a link.

Andrew Marsh
  • 2,032
  • 15
  • 14
15

Heads up for JavaScript users: all IDs become global variables under window.

<h1 id="foo">Foo Title</h1>

Just created the JS global:

window.foo

The value of window.foo will be the HTMLElement for the h1.

Unless you can guarantee all values used in id attributes are safe, you may prefer sticking to name:

<h1 name="foo">Foo Title</h1>
mikemaccana
  • 110,530
  • 99
  • 389
  • 494
  • 12
    The good news is that you can't overwrite the functions defined in `window`. For example, `
    ` won't overwrite the function `window.open`.
    – Flimm Jun 15 '15 at 10:52
  • 1
    Why should you guarantee that all values used in id attributes are safe? – Alex78191 Mar 22 '22 at 06:09
  • 1
    Article about this gotcha: https://css-tricks.com/named-element-ids-can-be-referenced-as-javascript-globals/ – Flimm Oct 06 '22 at 09:59
  • Thanks @Flimm - that's a good reference, I've included it as a link in the answer. – mikemaccana Oct 06 '22 at 10:25
9

ID method will not work on older browsers, anchor name method will be deprecated in newer HTML versions... I'd go with id.

Joannes Vermorel
  • 8,976
  • 12
  • 64
  • 104
Spikolynn
  • 4,067
  • 2
  • 37
  • 44
9

There's no semantic difference; the trend in the standards is toward the use of id rather than name. However, there are differences that may lead one to prefer name in some cases. The HTML 4.01 specification offers the following hints:

Use id or name? Authors should consider the following issues when deciding whether to use id or name for an anchor name:

  • The id attribute can act as more than just an anchor name (e.g., style sheet selector, processing identifier, etc.).
  • Some older user agents don't support anchors created with the id attribute.
  • The name attribute allows richer anchor names (with entities).
erickson
  • 265,237
  • 58
  • 395
  • 493
  • 14
    To be clear, when they say "older user agents" they mean REALLY old user agents. I wouldn't worry about that. – Eli Jun 25 '09 at 19:56
  • 1
    HTML5 allows “rich” IDs as well. Does anyone have version numbers of browsers with a market share larger than 0.1% that can’t handle id-anchored fragments? – Or is the dinosaur Netscape 4.7 actually the most spread one? – Robert Siemer Jul 29 '11 at 17:39
  • 7
    FWIW, I couldn't get `id` anchors to work in Safari for iOS 5, so it's not just browsers that were already "really old" in '09. I had to add `name`s to get my site to work properly on the iPad. This might have been fixed by now, I don't own any iOS 6 devices to check. – Daniel Saner Nov 21 '12 at 10:35
  • @DanielSaner really? so http://en.wikipedia.org/wiki/IPad#Applications doesn't work on your iPad? – Aprillion May 01 '13 at 12:26
  • I just checked again; it doesn't work with Safari on my iPad 2 with iOS 5. Your link always lands me at the top of the article. It does work with Opera though. I also tried it on my phone: `id` anchors don't work on the standard browser that comes with Gingerbread 2.3.7, but do in recent versions of Opera and Firefox. – Daniel Saner Jun 30 '13 at 20:46
  • 1
    @DanielSaner I used simulators to test Mobile Safari 5.02 & 5.1, and Android Browser 2.2 & 2.3, and the `id` anchors seem to work universally. If [this simple example](https://googledrive.com/host/0BxFi7p_MEU4ESFNnVENzR2pnVFE/anchor-test.html#p2) doesn't work on your mobile, I would check the device accessibility settings. (@deathApril Wikipedia mobile site has Javascript that effectively causes the URL fragment to be ignored.) – Stephen M. Harris Feb 07 '14 at 19:48
5

In html 5, the id="" attribute defines a unique identifier for an element, which is also an anchor for a fragment link. In previous html standards, the name="" attribute of the <a> element defines an anchor for a fragment link. I recommend something like:
<a name="foo" id="foo"></a><h1>Foo Title</h1>
Because support for the id="" attribute is a bit spotty(even though the latest releases of all major browsers support it, the releases that don't aren't more than a few years old[And it's best not to break something if there isn't a good reason to]). It's compatible, & it doesn't style whatever's in the link'd element, for the closing </a> is still outside the element, but it's still valid in all current standards.

Be sure that the name="" and id="" attributes of the <a> element are the same.

JustinCB
  • 513
  • 4
  • 12
  • 2
    Not everything that uses HTML is a browser. I'm using a Java library that displays information in a window using HTML. This is the only method that worked. It's the `name` attribute on an anchor tag that was required; placing attributes on an `hN` or `span` didn't work. – Mars Dec 22 '18 at 17:48
3

I have a web page consisting of a number of vertically stacked div containers, identical in format and differing only in serial number. I wanted to hide the name anchor at the top of each div, so the most economical solution turned out to be including the anchor as an id within the opening div tag, i.e,

<div id="[serial number]" class="topic_wrapper">
maximus
  • 165
  • 2
  • 7
2

The second sample assigns a unique ID to the element in question. This element can then be manipulated or accessed using DHTML.

The first one, on the other hand, sets a named location within the document, akin to a bookmark. Attached to an "anchor", it makes perfect sense.

Cerebrus
  • 25,615
  • 8
  • 56
  • 70
2

Just an observation about the markup The markup form in prior versions of HTML provided an anchor point. The markup forms in HTML5 using the id attribute, while mostly equivalent, require an element to identify, almost all of which are normally expected to contain content.

An empty span or div could be used, for instance, but this usage looks and smells degenerate.

One thought is to use the wbr element for this purpose. The wbr has an empty content model and simply declares that a line break is possible; this is still a slightly gratuitous use of a markup tag, but much less so than gratuitous document divisions or empty text spans.

jerseyboy
  • 1,298
  • 13
  • 13
  • That's a good point. More info on `wbr`: http://www.w3.org/TR/html-markup/wbr.html Use `` rather than `` – Luke Mar 19 '15 at 21:07
1

How about using name attribute for old browsers and id attribute to the new browsers. Both options will be used and fallback method will be implemented by default!!!

Umesh Bagalur
  • 311
  • 3
  • 11
0

It's not about support right now, because who would care in 2021 about older browsers than IE6?

Notice that Stackoverflow is still using anchor and 'name' attribute instead of 'name' attribute.

Why?

Because you:

  1. can't have two values for id on SINGLE element.

#sample
{
  color: red; 
}
<div id="postNr4245 sample"> Sample text </div>  <!-- coloring doesn't work -->

<a href="#postNr4245">Jump to Post</a> <!-- jumping doesn't work -->

Ok, you can say you can use 'class' for the sake of things like that.

It's true but if there is only single element that needs to be styled on your site then you shouldn't use class.

Why?

Because if you use CSS selector #idName and web browser finds that element it won't look further. It speeds loading of your website.

If you use .class it needs to look for every place of your DOM because web browser doesn't know if you had used .class once or more times.

  1. You can't start 'id' value with number if you want to style it in CSS or use JS to parse it. You can however start 'name' value with number.
Morfidon
  • 1,469
  • 1
  • 18
  • 34
-3

The whole "named anchor" concept uses the name attribute, by definition. You should just stick to using the name, but the ID attribute might be handy for some javascript situations.

As in the comments, you could always use both to hedge your bets.

GEOCHET
  • 21,119
  • 15
  • 74
  • 98
Alex Fort
  • 18,459
  • 5
  • 42
  • 51
  • When using both, are the id:s and names globally unique? as in, can I use the same string as both the id and the name? – Henrik Paul Jan 27 '09 at 19:13
  • You can, but some people think it's bad practice. – Alex Fort Jan 27 '09 at 19:18
  • 10
    According to the HTML specification, if both are present, name and id should be identical. It also says that names and ids are in the same namespace. The HTML validator service doesn't check for these, and I doubt browsers care, but they seem like good guidelines to follow anyway. – erickson Jan 27 '09 at 19:40
  • 4
    Reality redefined! – Robert Siemer Jul 29 '11 at 17:42