The Spanish forms of the numerals were called the _[h.]ur[=u]f al-[.g]ob[=a]r_, the [.g]ob[=a]r or dust numerals, as distinguished from the _[h.]ur[=u]f al-jumal_ or alphabetic numerals. Probably the latter, under the influence of the Syrians or Jews,[250] were also used by the Arabs. The significance of the term [.g]ob[=a]r is doubtless that these numerals were written on the dust abacus, this plan being distinct from the counter method of representing numbers. It is also worthy of note that Al-B[=i]r[=u]n[=i] states that the Hindus often performed numerical computations in the sand. The term is found as early as c. 950, in the verses of an anonymous writer of Kairw[=a]n, in Tunis, in which the author speaks of one of his works on [.g]ob[=a]r calculation;[251] and, much later, the Arab writer Ab[=u] Bekr Mo[h.]ammed ibn `Abdall[=a]h, surnamed al-[H.]a[s.][s.][=a]r {66} (the arithmetician), wrote a work of which the second chapter was "On the dust figures."[252] The [.g]ob[=a]r numerals themselves were first made known to modern scholars by Silvestre de Sacy, who discovered them in an Arabic manuscript from the library of the ancient abbey of St.-Germain-des-Pres.[253] The system has nine characters, but no zero. A dot above a character indicates tens, two dots hundreds, and so on, [5 with dot] meaning 50, and [5 with 3 dots] meaning 5000. It has been suggested that possibly these dots, sprinkled like dust above the numerals, gave rise to the word _[.g]ob[=a]r_,[254] but this is not at all probable. This system of dots is found in Persia at a much later date with numerals quite like the modern Arabic;[255] but that it was used at all is significant, for it is hardly likely that the western system would go back to Persia, when the perfected Hindu one was near at hand. At first sight there would seem to be some reason for believing that this feature of the [.g]ob[=a]r system was of {67} Arabic origin, and that the present zero of these people,[256] the dot, was derived from it. It was entirely natural that the Semitic people generally should have adopted such a scheme, since their diacritical marks would suggest it, not to speak of the possible influence of the Greek accents in the Hellenic number system. When we consider, however, that the dot is found for zero in the Bakh[s.][=a]l[=i] manuscript,[257] and that it was used in subscript form in the _Kit[=a]b al-Fihrist_[258] in the tenth century, and as late as the sixteenth century,[259] although in this case probably under Arabic influence, we are forced to believe that this form may also have been of Hindu origin.